Post on 30-Jun-2020
transcript
Dis cus si on Paper No. 12-037
The Competitive Effects of Firm ExitEvidence from the U.S. Airline Industry
Kai Hüschelrath and Kathrin Müller
Dis cus si on Paper No. 12-037
The Competitive Effects of Firm ExitEvidence from the U.S. Airline Industry
Kai Hüschelrath and Kathrin Müller
Download this ZEW Discussion Paper from our ftp server:
http://ftp.zew.de/pub/zew-docs/dp/dp12037.pdf
Die Dis cus si on Pape rs die nen einer mög lichst schnel len Ver brei tung von neue ren For schungs arbei ten des ZEW. Die Bei trä ge lie gen in allei ni ger Ver ant wor tung
der Auto ren und stel len nicht not wen di ger wei se die Mei nung des ZEW dar.
Dis cus si on Papers are inten ded to make results of ZEW research prompt ly avai la ble to other eco no mists in order to encou ra ge dis cus si on and sug gesti ons for revi si ons. The aut hors are sole ly
respon si ble for the con tents which do not neces sa ri ly repre sent the opi ni on of the ZEW.
Non-technical summary
In the last decade, the domestic U.S. airline industry has experienced a substantial
consolidation trend. In addition to a number of high level mergers such as American Airlines
– Trans World Airlines (2001), America West – US Airways (2005) and Delta Air Lines –
Northwest Airlines (2009), several smaller carriers such as National Airlines (2002),
Independence Air (2006) and ATA Airlines (2008) had to leave the industry.
Despite this high relevance of firm exits for the recent development of the domestic U.S.
airline industry, empirical evidence on the effects of these consolidations is rare. Studies
focusing on the market impact of liquidations do not exist to the best of our knowledge and
the existing studies on the competitive effects of airline mergers almost exclusively stem from
the 1980s and focus on the specific case of a largely overlapping route network of the
merging parties (due to a shared hub). However, such a network structure is rather uncommon
in recent mergers and therefore raises the demand for both a new conceptual framework for
investigating firm exits in the airline industry and a corresponding new empirical analysis of
the effects of such firm exits.
Against this background, we study the competitive effects of five liquidations and six mergers
in the domestic U.S. airline industry between 1995 and 2010. Applying fixed effects
regression models we find that route exits due to liquidation lead to substantially larger and
permanent price increases (of about 12 percent) than merger-related exits. Within the merger
category, our analysis reveals that prices on overlapping routes increase by about 6 percent in
the short run, whereas a simple merger-induced switch of the operating carrier causes a
significant price increase of about 3 percent. Accordingly, we observe large reductions in
quantity for route exits caused by firm liquidations and moderate reductions in case of
merger-related exits. Entry-inducing effects of firm exits are found particularly for
liquidations and smaller mergers.
Das Wichtigste in Kürze
Im Laufe des vergangenen Jahrzehnts war in der US-amerikanischen Luftverkehrsindustrie
ein substanzieller Konsolidierungstrend zu beobachten. Neben einigen größeren Fusionen wie
beispielsweise American Airlines – Trans World Airlines (2001), America West – US
Airways (2005) und Delta Air Lines – Northwest Airlines (2009) gingen einige kleinere
Fluggesellschaften wie National Airlines (2002), Independence Air (2006) und ATA Airlines
(2008) in Konkurs und verließen die Industrie.
Trotz dieser festgestellten hohen Relevanz von Firmenaustritten für die jüngeren
Entwicklungen der US-amerikanischen Luftverkehrsindustrie sind empirische
Untersuchungen zu den ökonomischen Effekten dieser Konsolidierungen sehr spärlich gesät.
Während Studien mit einem Fokus auf konkursbedingte Firmenaustritte nach unserem
Kenntnisstand gar nicht existieren, konzentrierten sich fast alle existierenden Studien zu den
wettbewerblichen Effekten von Fusionen zwischen Fluggesellschaften auf Fälle aus den
1980er Jahren. Diese Fusionen waren allerdings gekennzeichnet von einem stark
überlappenden Luftverkehrsnetz der fusionieren Parteien (bedingt durch einen gemeinsamen
Hub-Flughafen); ein Merkmal, das aktuellere Fusionen nicht aufweisen. Es liegt daher nahe,
nicht nur ein neues konzeptionelles Grundgerüst für eine Analyse der wettbewerblichen
Effekte solcher Firmenaustritte zu entwickeln, sondern dieses in der Folge mit einer
empirischen Analyse von konkurs- und fusionsbedingten Firmenaustritten zu kombinieren.
Vor diesem Hintergrund untersuchen wir die wettbewerblichen Effekte von fünf Konkursen
und sechs Fusionen im US-amerikanischen Luftverkehr zwischen 1995 und 2010. Im Zuge
der Anwendung verschiedener Paneldatenmodelle mit fixen Effekten stellen wir fest, dass
konkursbedingte Marktaustritte zu substanziell höheren und permanenten Preisanstiegen (von
durchschnittlich 12 Prozent) führen als fusionsbedingte Austritte. Innerhalb der Kategorie der
fusionsbedingten Marktaustritte zeigt unsere empirische Analyse, dass auf überlappenden
Streckenmärkten die Preise durchschnittlich um 6 Prozent in der kurzen Frist ansteigen,
während für Routen, auf denen nur ein fusionsbedingter Wechsel der Fluggesellschaft
stattfindet, Preisanstiege von durchschnittlich 3 Prozent zu beobachten sind. Im Einklang mit
diesen Ergebnissen finden wir eine große Reduktion der angebotenen Kapazitäten im Falle
von Konkursen und einen eher moderaten Rückgang im Falle von Fusionen. Verstärkter
Markteintritt nach den entsprechenden Firmenaustritten kann insbesondere bei Konkursen
und Fusionen von kleineren Fluggesellschaften festgestellt werden.
THE COMPETITIVE EFFECTS OF FIRM EXIT
EVIDENCE FROM THE U.S. AIRLINE INDUSTRY
Kai Hüschelrath and Kathrin Müller
April 2012
Abstract We study the competitive effects of five liquidations and six mergers in the domestic U.S. airline industry between 1995 and 2010. Applying fixed effects regression models we find that route exits due to liquidation lead to substantially larger price increases than merger-related exits. Within the merger category, our analysis reveals significant price increases on all affected routes immediately after the exit events. In the medium and long-run, however, realized merger efficiencies and entry-inducing effects are found to be strong enough to drive prices down to pre-exit levels.
Keywords Airline industry, exit, liquidation, merger, efficiencies, entry-inducing effects
JEL Class L40, L93
Head, Competition and Regulation Research Group, ZEW Centre for European Economic Research, P.O. Box 10 34 43, D-68034 Mannheim, Germany, E-mail: hueschelrath@zew.de; Coordinator, MaCCI Mannheim Centre for Competition and Innovation; Assistant Professor for Industrial Organization and Competitive Strategy, WHU Otto Beisheim School of Management, Burgplatz 2, 56179 Vallendar, Germany.
Researcher, Competition and Regulation Research Group, ZEW Centre for European Economic Research and MaCCI Mannheim Centre for Competition and Innovation, P.O. Box 10 34 43, D-68034 Mannheim, Germany, E-mail: kathrin.mueller@zew.de. We are indebted to Volodymyr Bilotkach, Jozsef Molnar and session participants at the 2012 International Industrial Organization Conference in Arlington for valuable comments on previous versions of the paper. The usual disclaimer applies.
1
1 INTRODUCTION
The benefits of competition and innovation are largely ensured by both market entry and
market exit. Market entry plays a key role as an equilibrium force – which competes away
excess profits to an equilibrium level – and as a disequilibrium force – which propels the
industry from one equilibrium state to another due to the introduction and diffusion of
innovations (see Geroski, 1991, 1995). Market exit is considered a key instrument to sanction
unprofitable product and service ideas thereby renewing the industry population through a
process of ‘creative destruction’ (Schumpeter, 1942). Only the close interaction of market
entry of innovative and/or more efficient new firms and the corresponding decline and market
exit of less innovative and/or less efficient incumbent firms through either merger or
liquidation can guarantee dynamically efficient markets.
The U.S. airline industry has experienced many firm entries and exits since its deregulation
in 1978. For example, following a transition period in the years after deregulation with a
moderate number of in sum 13 firm entries and 6 firm exits, the substantial growth period
from 1982 to 1984 witnessed the entry of 31 new interstate airlines – excluding regional
carriers – compared to only 15 exits either through merger or liquidation. However, the
subsequent shake-out period from 1985 to 1987 showed roughly inverted characteristics with
only 16 additional entries compared to 38 exits (mostly through liquidation).1
Although the number of mergers and liquidations in the last two decades has been
substantially lower than in the first shake-out phase of the liberalized U.S. airline industry,
both types of firm exit continue to have a substantial impact on the industry. This is
particularly true for large mergers such as American Airlines – Trans World Airlines (2001),
America West – US Airways (2005) and Delta Air Lines – Northwest Airlines (2009) but also
for larger liquidations such as National Airlines (2002), Independence Air (2006) and ATA
Airlines (2008).
Despite this continuing relevance of firm exits in the U.S. airline industry, recent empirical
evidence is rare. Empirical studies on the effects of liquidations do not exist to the best of our
knowledge and the existing studies on the competitive effects of airline mergers almost
exclusively stem from the 1980s and focus on the specific case of a largely overlapping route
1 The data stems from Borenstein and Rose (2008). Interestingly, the authors report that out of the group of 44 (interstate) carriers that entered the U.S. airline industry between 1979 and 1984, only 7 operated in 1990 and only two remain in operation today (Southwest Airlines and America West (using the US Airways brand)).
2
network of the merging parties (due to a shared hub). However, such a network structure is
rather uncommon in recent mergers and therefore raises the demand for both a new
conceptual framework for investigating firm exits in the airline industry and a corresponding
new empirical analysis of the effects of such exits.
Against this background, we study the competitive effects of five liquidations and six
mergers in the domestic U.S. airline industry between 1995 and 2010. Applying fixed effects
regression models we find that route exits due to liquidation lead to substantially larger and
permanent price increases (of about 12 percent) than merger-related exits. Within the merger
category, our analysis reveals that prices on overlapping routes increase by about 6 percent in
the short run, whereas a simple merger-induced switch of the operating carrier causes a
significant price increase of about 3 percent. Accordingly, we observe large reductions in
quantity for route exits caused by firm liquidations and moderate reductions in case of
merger-related exits. Entry-inducing effects of firm exits are found particularly for
liquidations and smaller mergers. Our findings have important implications for unilateral
effects analysis as part of horizontal merger assessments.
The paper is structured as follows. The following second section provides a conceptual
framework for the analysis of the competitive effects of firm exit in the airline industry,
followed by the discussion of descriptive evidence for the U.S. airline industry in the third
section. The fourth section presents our empirical analysis and provides a discussion of the
results and their policy implications. The fifth section concludes the paper with a summary of
the key results.
2 THE COMPETITIVE EFFECTS OF FIRM EXIT IN THE AIRLINE
INDUSTRY – A CONCEPTUAL FRAMEWORK
Market exit can be assessed on two different aggregation levels: exit of entire companies due
to either liquidation or merger (so-called firm exits) and single market exit decisions of still
operating companies for strategic reasons such as lack of profitability2 or – in case of the
airline industry - network reorganization (so-called operational exits). Firm exits differ from
operational exits by the fact that the former typically cause multiple market exits at one
particular point in time. Furthermore, while operational exists can typically be reversed if,
2 Several reasons for an unprofitable route presence are conceivable: incumbent(s) reaction(s) to entry, wrongly estimated demand (O&D and/or connecting traffic), insufficient growth potential, increases in passenger facility or airport charges, macroeconomic demand shocks, input cost increases etc.
3
e.g., market conditions change, firm exits are ultimate thereby reducing (actual and potential)
competition on a permanent basis.
The economic effects of the two distinctive forms of market exit can be analyzed from at
least two perspectives: ‘general economic effects’ and ‘competitive effects’. The former
category investigates the consequences of market exit on general economic factors such as
employment or (regional) economic growth. Any study of these general economic effects
must look beyond the level of the respective firm and its product markets and has to include
important knock-on effects of market exit on, e.g., the airport, other aviation-related service
industries or spillovers to the general economic growth in the respective region. For example,
if an airline decides to leave a certain hub airport, either due to liquidation (i.e., firm exit) or
to network reorganization (i.e., operational exit), it is very likely that the respective airport
and other aviation-related service industries will lose business. Furthermore, the entire region
might face a reduction in attractiveness due to the lower quality of airline connections.
Complementary to an analysis of the effects of exit on the general economic level, an
assessment of the competitive effects of exit is a compulsory part of an entire analysis of the
economic consequences of exit. Generally, such an assessment investigates the effects of
market exit on competition in these markets. Particularly interesting objects of investigation
are the effects on average prices, demand, capacity and quality. For example, if before exit,
two airlines were competing fiercely in a certain non-stop market and one of the competitors
finally has to exit the market due to liquidation, it becomes likely that the remaining carrier
will use this opportunity to increase price. However, in the medium and long run, market
entry by other (more efficient) airlines might become attractive, i.e., firm exit might create so-
called entry-inducing effects (see Werden and Froeb (1998) for a seminal paper3) suggesting
that prices might increase immediately after exit but exhibit a downward trend in the medium
and longer run. As a consequence, a study on the competitive effects of exit should not be
constrained to an analysis of the short-term effects on price but has to extend its perspective to
both a larger observation window and the inclusion and interpretation of further competition
variables such as passengers and departures.
3 Werden and Froeb (1998) investigate the role of entry-inducing effects in antitrust policy. Based on mergers in simple Cournot and Bertrand industries, they find that firms only have an incentive to merge if (a) they expect significant efficiencies generated from the merger, or (b) they are aware of substantial entry barriers which allow them to charge supracompetitive profits post-merger. They conclude that antitrust authorities should be rather skeptical with respect to the power of entry to prevent (or reverse) anticompetitive effects of horizontal mergers.
4
In the remainder of this paper, we concentrate on the competitive effects of firm exit. The
firm exit category is subdivided further into multiple market exits due to liquidation and
multiple market exits due to merger. In our empirical investigation, we include ‘operational
exits’ as third category in order to allow for a direct comparison between the effects of firm
exits and single market exits. For a detailed discussion on the competitive effects of firm exit,
we draw on a simple airline network with two separate airlines 1 and 2 operating hubs H1 and
H2, respectively (see Figure 1).
Figure 1: A simple network with two airlines Source: own figure
The two hubs are connected by services of both carriers (route H1H2) while the respective
spokes are only served by the respective hub airline, i.e., routes AH1, BH1 and CH1 by airline
1 and H2D, H2E and H2F by airline 2. In the following sub-sections, we discuss the
competitive effects of a liquidation, i.e., airline 1 disappears from the market, and the
competitive effects of a merger, i.e., airline 2 acquires airline 1 and continues operating the
entire network.
2.1 FIRM EXIT THROUGH LIQUIDATION
The bankruptcy laws in many countries allow two different forms of bankruptcy: the attempt
of reorganization (and a potential ‘emergence’ from bankruptcy) or a process of liquidation
(which typically leads to the exit of the respective firm). As the focus in this paper is on the
effects of firm exit, we concentrate on those bankruptcies which lead to the ultimate market
exit (i.e., liquidation) of the respective firm.
When studying the competitive effects of liquidation exits on prices and quantity, basic
oligopoly theory allows the derivation of several key relationships. In the short run, at least
two separate arguments speak for significant price increases. First, the substantial reduction in
capacity due to exit of one competitor is expected to cause substantial price increases and
corresponding reductions in quantities. Second, pre-exit competition on the respective routes
5
might have been fierce (as at least one carrier was fighting for life) suggesting price increases
‘up to the usual competitive level’ post-exit.4 Referring to the simple airline network defined
above, the described competitive effects are expected on all routes operated by airline 1. If no
other competitors are operating on the respective spoke routes AH1, BH1 and CH1, (non-stop)
services are terminated completely.
In the medium and long run, however, the anticipated effects on the key variables are less
clear. On the one hand, existing carriers might find it profitable to expand capacity in the
respective markets, thereby putting pressure on price. On the other hand, other airlines might
have incentives to enter the respective routes (the hub-to-hub as well as the spoke routes)
thereby increasing competition and triggering a downward trend in price. However, the
respective possibilities of entry depend on both profit expectations on the one hand and the
size of (structural and/or strategic) barriers to entry on the other hand leaving the direction and
size of the medium- and long run effects of firm exit through liquidation unclear. For
example, large scale entry into H1 might not be too attractive given the fact that another
airline just exited the respective airport. Furthermore, entry into H2 might be difficult
(technically and economically) as soon as airline 2 has a dominant position at this hub.
2.2 FIRM EXIT THROUGH MERGER
Mergers and acquisitions (‘mergers’ in the following) are another form of firm exit. In order
to derive hypotheses on the effects of mergers on competition, we have to introduce a
separation of possibly affected routes. So-called overlapping routes are characterized by the
presence of both merging carriers before the merger, i.e., they are direct competitors in the
respective non-stop markets (route H1H2 in Figure 15). In contrast, we introduce a second
category of routes for which possible merger effects cannot be ruled out: the so-called
switching routes. Switching routes are operated by the junior merger partner ex-ante but are
taken over by the lead merger partner as a consequence of the merger (routes AH1, BH1 and
CH1 in Figure 1 if airline 2 acquires airline 1).6
4 From this perspective, liquidation has the potential to realize ‘liquidation efficiencies’ in the sense that the market exit of one carrier allows the other carriers to earn a reasonable return on investment and to continue serving the respective market.
5 Although Figure 1 only shows one overlapping route (H1H2), overlaps in real airline networks are not restricted to hub-to-hub operations but can also be observed in other constellations such as route CH2 or route AD. Our empirical analysis below includes all different types of overlapping routes.
6 In general, a key determinant of the possibilities to increase price post-merger are the network structures pre-merger. If the networks are largely complementary as assumed in the simple network structure in Figure 1, theory and empirical evidence suggest that anticompetitive effects are only likely on the (few) routes on which both merging parties are operating before the merger. However, if the degree of route overlap is
6
Focusing on overlapping routes first, in the short run, the capacity of the junior carrier is
unlikely to be removed entirely from the market. However, mergers might lead to an increase
in market power thereby incentivizing the merged entity to reduce quantity and increase price.
Ceteris paribus, price increases immediately after the merger should be less pronounced than
price increases after liquidation exits. In the medium and long run, the notional effects are
again less clear. On the one hand, capacity expansions of existing competitors or entry by
potential competitors might constrain the market power of the merged parties. On the other
hand, barriers to entry such as hub dominance might reduce the threat of entry substantially
thereby allowing the merged parties to permanently increase prices. Comparing liquidation-
related exits and merger-related exits, it can be expected that capacity increases through entry
are more likely after liquidations given the substantial loss of capacity due to the exit of one
carrier. Merger-related exists, however, are more likely to put the merging parties in a
stronger position on the respective overlapping routes suggesting a dissuading effect on the
entry incentives of potential competitors.
Turning from overlapping routes to switching routes, an initial assessment of this route
type might suggest that competitive effects are rather unlikely, especially because the number
of competitors on these routes is unaffected by the merger. However, a closer look reveals
that the merger-induced change in the operator might contain several possibilities for price
reactions. First, a change in ownership might cause changes in pricing and other strategic
variables possibly triggering significant price changes post-merger. Second, the merger might
have an impact on the quality of the merged product. For example, the merger of two
complementary networks creates additional travel possibilities for the customers of both
airlines thereby increasing quality (and possibly justifying price increases). Last but not least,
the merger increases multimarket contact among the remaining airlines in the industry and
might therefore ease the realization of (tacitly) collusive outcomes.
Although horizontal mergers unavoidably raise market concentration, a countervailing
force to market power-induced price increases are possible merger efficiencies (see
Williamson (1968) for a seminal paper). Such efficiencies might come in the form of both
savings in marginal costs and fixed costs and are likely to be realized in the medium and long
run. Merger efficiencies might create incentives to pass on a significant share down to the
consumers through price reductions post-merger. Again, it is difficult ex-ante to derive clear
substantial, e.g., due to hub operations of both airlines at the same airport, anticompetitive effects on a large number of non-stop connections are much more likely.
7
hypotheses in this respect. In general, efficiencies can partly become effective within the
entire merged airline network (e.g., through a cheaper sourcing of input goods for the merged
entity) and might partly be attributed to certain improvements of particular hub presences etc.
In other words, efficiencies can play a role in both overlapping and switching routes. A key
problem in an assessment of merger efficiencies is the unknown time-frame of their
realization. While marginal cost savings might be realized relatively quickly after the merger,
fixed costs savings might take substantially longer (and are therefore difficult to isolate
empirically from other effects that might influence the price and cost levels of the respective
airline).
3 FIRM EXIT IN THE U.S. AIRLINE INDUSTRY
Based on the general discussion of firm exit in the preceding section, this section investigates
the issue specifically for the U.S. airline industry. Differentiating between firm exits and
operational exits, Figure 2 shows the number of exits in non-stop domestic U.S. airline
markets between the third quarter of 1995 and the first quarter of 2010.
0
20
40
60
80
100
120
140
1995‐3
1996‐1
1996‐3
1997‐1
1997‐3
1998‐1
1998‐3
1999‐1
1999‐3
2000‐1
2000‐3
2001‐1
2001‐3
2002‐1
2002‐3
2003‐1
2003‐3
2004‐1
2004‐3
2005‐1
2005‐3
2006‐1
2006‐3
2007‐1
2007‐3
2008‐1
2008‐3
2009‐1
2009‐3
2010‐1
number of exits
merger exits liquidation exits operational exits
Figure 2: Exits from non-stop domestic U.S. airline markets (3rd quarter 1995 – 1st quarter 2010)7
Source: DOT T-100 Segment Data, authors’ calculations.
The quarter of exit is defined as the first quarter after an airline provided the last non-stop
service between two airports. In sum, the observation period shows that operational exits
fluctuate substantially between the years with peaks after the 9/11 attacks and the beginning
7 The graph shows the (aggregated) exit activities of the following 27 major U.S. carriers: AA, AS, B6, BF, CO, DH, DL, F9, FF, FL, G4, HA, HP, J7, N7, NJ, NK, NW, QQ, SY, TW, TZ, UA, US, VX, WN, YX (1).
8
of the economic crisis in 2008. The figure further reveals that six larger mergers and five8
larger liquidations took place between 1999 and 2010. The by far largest number of route
exits due to firm exit was triggered by the Delta Air Lines – Northwest Airlines merger which
materialized in the first quarter of 2010. The by far largest liquidation case was Independence
Air in the first quarter of 2006. In general, most liquidations and mergers create significant
‘shocks’ in the exit statistics in the respective quarters. The following sections will take a
closer look at firm exits through liquidation and merger in the U.S. airline industry between
1995 and 2010.
3.1 FIRM EXIT THROUGH LIQUIDATION
In the United States, Chapter 7 of the Title 11 of the Bankruptcy Code governs the process of
liquidation while Chapters 11 and 13 govern the process of reorganization of a debtor in
bankruptcy. While Chapter 7 is the usual form of bankruptcy in the United States, most airline
bankruptcies refer to Chapter 11. However, although most airlines successfully went through
the process of reorganization and exited Chapter 11 at some point, several carriers eventually
were liquidated. In fact, most liquidations of U.S. carriers were carried out under Chapter 11
and not under Chapter 7. For the period from 1995 to 2010, we identified five liquidations of
larger U.S. airlines (see Table 1).
Table 1: Liquidations of larger U.S. airlines between 1995 and 2010
Airline IATA Code
Year of firm entry
Quarter of firm
exit
Number of domestic non-stop routes
operated (quarter before/of firm exit)
Number of domestic passengers
transported (quarter before/of firm exit)
Share of domestic passengers
transported (quarter before/of firm exit)
ATA Airlines TZ 1986 2008-2 15 372,412 2.35‰
Independence Air DH 2004 2006-1 44 1,013,483 6.47‰
Vanguard Airlines NJ 1994 2002-4 17 199,747 1.42‰
National Airlines N7 1999 2002-4 10 612,514 4.35‰
Tower Air FF 1983 2000-2 6 121,850 0.87‰
Sources: Airlines for America (http://www.airlines.org/Pages/U.S.-Airline-Bankruptcies-and-Service-Cessations.aspx), U.S. DOT, T-100 Domestic Segment Data, authors’ calculations.
As shown in Table 1, all liquidation cases were relatively small, reflected in measures such as
the number of non-stop routes operated in the quarter before exit or the share of passengers
transported compared to the number of passengers in the entire domestic U.S airline industry.
However, despite the small size of all liquidation cases compared to the entire national
market, it would be superficial to automatically conclude the irrelevance of an economic
analysis. As all exiting airlines had a particular geographic focus, the impact on this particular
8 National Airlines and Vanguard Airlines were liquidated in the same quarter (2002-4).
9
set of routes in this particular region can be quite substantial and therefore justifies a detailed
investigation.
Existing empirical research on the competitive effects of U.S. airline bankruptcies
concentrates solely on cases in which the respective airlines were not liquidated but entered a
temporary phase of reorganization. In a seminal paper, Borenstein and Rose (1995)
investigate the effects of bankruptcy filings by seven U.S carriers on market conduct. They
find that carriers lowered fares by 5 to 6 percent before entering bankruptcy, however,
refrained from further fare cuts after entering bankruptcy protection. Furthermore, the study
reveals that the fare reductions are only observable in the short term and are not followed by
route-level competitors. Ciliberto and Schenone (2008) not only confirm this last result, but
especially find that the observed changes in prices must be generated by a reduction in
capacity by the bankrupt firms and not by other explanations such as cost reductions or
changes in strategic behavior. In a sense, these results conflict with earlier findings by Busse
(2002). She uses data on 14 major airlines between 1985 and 1992 and concludes that firms in
worse financial condition are more likely to start price wars. Last but not least, Borenstein and
Rose (2003) specifically investigate the impact of bankruptcy on airline service levels and
find that bankruptcies reduce service levels (as measured in the ‘number of nonstop flights’
and the ‘number of destinations served on a non-stop basis’) at some airports. The estimated
effect is found to be the greatest for midsize airports (i.e., airports between 100 and 400
flights per day).
Although the studies on the competitive effects of bankruptcy reorganization come to
interesting and policy-relevant conclusions, economic theory suggests that the competitive
effects of airline liquidations are different. As argued above, liquidation causes multiple route
exits at one particular point in time which cannot be reversed thereby reducing (actual and
potential) competition on a permanent basis. Given the absence of studies which empirically
investigate the direction and size of the competitive effects of firm liquidations, we provide
such an analysis in the fourth section below.
10
3.2 FIRM EXIT THROUGH MERGER
In the history of the U.S. airline industry, both types of firm exit, liquidations and mergers,
have often been interpreted as closely interrelated in the sense that (especially large) bankrupt
airlines were acquired by a competitor to basically avoid liquidation.
Table 2: Mergers and acquisitions of larger U.S. airlines between 1995 and 2010
Merging parties Closed Quarter of
merger exits
Resulting carrier
Number of junior partner’s domestic non-stop route exits due to merger
Junior partner’s domestic
passengers (quarter before/of
merger)
Share of junior partner’s domestic
passengers (quarter before/of
merger)
Delta Air Lines / Northwest Airlines
12/31/2009 2010-1 Delta
Air Lines 148 7,220,155 4.770%
Frontier Airlines / Midwest Airlines
7/31/2009 2009-4 Frontier Airlines
11 529,666 0.321%
US Airways / America West Airlines
9/27/2005 2005-4 US
Airways 91 6,859,074 4.059%
American Airlines / TWA
4/9/2001 2001-3 American Airlines
85 6,152,064 3.924%
American Airlines / Reno Air
2/1/1999 1999-2 American Airlines
24 1,247,481 0.928%
AirTran Airways / Valujet
11/17/1997 1998-1 AirTran Airways
27 718,592 0.535%
Sources: Airlines for America (http://www.airlines.org/Pages/U.S.-Airline-Mergers-and-Acquisitions.aspx), U.S. DOT, T-100 Domestic Segment Data, authors’ calculations.
As shown in Table 2, the observation period has experienced six mergers between larger
U.S. airlines, including (at least) three ‘failing firms’ (Trans World Airlines, US Airways, and
Valujet). As reflected in the correspondingly high numbers of non-stop route exits due to the
merger or the share of passengers in the quarter before the merger (relative to all domestic
U.S. airline passengers), three mergers have been particularly large transactions – American
Airlines – Trans World Airlines (2001), America West – US Airways (2005) and Delta Air
Lines – Northwest Airlines (2010) – while the remaining three transactions were of
significantly smaller size. However, as already argued for the liquidation cases above, it
would be superficial to automatically ignore these smaller transactions, e.g., due to possibly
substantial effects on the affected routes.
Existing empirical research on the competitive effects of U.S. airline mergers largely refers
to the late 1980s.9 On the one hand, this period was characterized by a substantial industry
9 An incomplete list of empirical articles focusing on the competitive effects of U.S. airline mergers includes Beutel and McBride (1992), Borenstein (1990), Brueckner et al. (1992), Butler and Huston (1989), Jordan (1988), Kim and Singal (1993), Morrison (1996) and Werden et al. (1991). Furthermore, an event study
11
consolidation leading to a large number of mergers as possible study objects. On the other
hand, the Department of Justice followed a laissez-faire approach to antitrust policy at that
time – strongly influenced by the theory of contestable markets by Baumol et al. (1982) – and
leading to the approval of basically all merger proposals independent of their potential for
anticompetitive effects.
Two U.S. airline mergers – both completed in 1986 – experienced a particularly detailed
ex-post investigation of their competitive effects: Northwest Airlines – Republic Airlines
(NW-RC) and Trans World Airlines – Ozark Airlines (TW-OZ). Both mergers involved a
shared major hub airport and therefore led to substantial increases in market power post-
merger. In a first influential paper, Werden et al. (1991) investigate the price and output
effects of the two mergers at their respective hub airports and find yield increases of about
5.6 percent and service decreases of about 23.7 percent for the NW-RC merger. Yield
increases (1.5 percent) and service decreases (16.2 percent) were somewhat smaller for the
TW-OZ merger. Borenstein (1990) analyzes the effects of the same two mergers at their hub
airports and finds evidence for price increases for the NW-RC merger of about 9.5 percent in
total (with about 6.7 percent price increases if other airlines remain as route competitors and
about 22.5 percent if the merger led to a monopoly route). For the TW-OZ merger, however,
his analysis resulted in largely insignificant results with the exception of a significant price
decrease of about 12.3 percent on monopoly routes which were operated by TW or NZ before
the merger.10 Interestingly, Borenstein’s analysis therefore showed that the mergers had an
impact “not just on routes that both airlines had served prior to the merger, but also on routes
where only one of the two merger partners competed with another airline or operated without
active competition” (Borenstein (1990), p. 404). He explains this finding by the possibilities
to reduce the threat of potential competition due to increased airport dominance.
Borenstein’s key result of merger effects on routes in which only one of the merging
carriers was active pre-merger is confirmed in studies by Kwoka and Shumilkina (2010) and
Kim and Singal (1993). While Kwoka and Shumilkina (2010) also analyze a single merger
(USAir and Piedmont in 1987) and find that prices rise by 5 to 6 percent on routes which were
only served by one of the merging carriers and the other was a potential entrant, Kim and
approach is followed in studies by Knapp (1990), McGuckin et al. (1992), Singal (1996) and Slovin et al. (1991).
10 It is important to note here that the observed price decrease is rather unexpected and might be explained by a general period of low demand at TWA’s St. Louis hub. For the NW-RC merger, Borenstein (1990) finds significant price increases of about 6 percent for NW or RC routes in which (a) competitor(s) remain after the merger and price increases of about 12 percent for NW or RC routes which became a monopoly post-merger.
12
Singal (1993) analyze the effects of fourteen U.S. airline mergers between 1985 and 1988 and
find that relative fares on the merging firms’ routes rose by about 9.4 percent. Significant
price increases were particularly found on routes in which the merging parties did not
compete (directly) prior to the merger. They explain this observation by an increase in multi-
market contact triggered by the merger. Furthermore, the authors identified a substantial
difference in the behavior of ‘mergers including a failing firm’ and ‘mergers without a failing
firm’. Fares of failing airlines were found to be much lower on average before the merger,
providing an explanation for the substantially larger price increases after the merger compared
to cases of mergers between ‘healthy’ firms.
Partly due to the substantial reduction in merger activity in the 1990s and 2000s, existing
research on the competitive effects of more recent U.S. airline mergers is very limited. From
an ex-post perspective, Bilotkach (2011) investigates the America West – US Airways merger
with a particular focus on its implications for multimarket contact (MMC). He finds that the
merger changed the way that the airlines take into account the extent of MMC when making
strategic choices as to frequency of service. From an ex-ante perspective, constant rumours of
possible mega-mergers led to several policy studies on the possible effects of such mergers
(see, e.g., U.S. General Accounting Office, 2001, U.S. Government Accountability Office,
2010). However, academic contributions are restricted to a research paper by Benkard et al.
(2010) in which the authors simulate the dynamic effects of three proposed horizontal U.S.
airline mergers. Using data for 2003-2008, they find that a merger between two major hub
carriers leads to increased entry by both other hub carriers and low cost carriers thereby
offsetting some of the initial concentrating effects of the merger.
The existence of both a significant number of liquidations and mergers in recent years
demands a detailed econometric investigation of the competitive effects of these firm exits.
The following section provides such an analysis for the route exits caused by five liquidations
and six mergers which took place in the domestic U.S. airline industry between 1998 and
2010.
4 EMPIRICAL ANALYSIS
In this section, we present our empirical analysis. While Section 4.1 describes the
construction of the dataset, Section 4.2 specifies our empirical approach and Section 4.3
provides the descriptive statistics. Subsequently, Section 4.4 concentrates on the presentation
13
and interpretation of our main empirical results followed by the discussion of important
policy implications in Section 4.5.
4.1 CONSTRUCTION OF THE DATASET
Our dataset was constructed by collecting and merging data from several sources. We use
airline traffic data for the years from 1995 to 2011 from the U.S. DOT T-100 Domestic
Segment database. This data contains monthly domestic non-stop segment data reported by
U.S. air carriers when both origin and destination airports are located within the boundaries of
the United States and its territories. We use T-100's information on origin, destination, non-
stop distance, available capacity, number of departures, and number of passengers to
construct a quarterly panel data-set of non-directional non-stop route airport-pair markets. We
drop airline-route observations with less than 12 quarterly departures and airline-route
observations which were only served one quarter between 1995 and 2011. In addition, we use
fare data from the U.S. DOT DB1B Market Origin and Destination Survey to enrich the
constructed panel dataset with quarterly route-level fare data. In detail, the construction of the
dataset can be subdivided into the following three subsequent steps.
In the first step, we identify all route exits of the 27 largest U.S. carriers11 which have been
taken place between the 3rd quarter of 1995 and the 1st quarter of 2010. The quarter of exit is
defined as the quarter following the quarter of the last occurrence of an airline-route
observation in the dataset. Liquidation exits are all exits of the respective carrier which
occurred in the quarter of firm liquidation (see Table 1 above). Merger-related exits are
assumed to have taken place in the quarter after the closure of the merger transaction (see
Table 2 above).
In the second step, we keep all non-stop routes which were subject to at least one exit
(operational exit, liquidation exit or merger exit) and which are still served by another carrier
after the exit of the respective carrier.12 If multiple exits on a certain non-stop route were
observed over time, the route was duplicated. For each exit, we keep the eight quarters before
and the eight quarters after the exit event to assess the effects of an exit using a ‘before-and-
after’ approach. We drop all routes for which we have less than seven observations before and
seven observations after exit.
11 See footnote 7 above for a list of these 27 major U.S. carriers. 12 If no non-stop service is provided after exit, route level effects cannot be observed.
14
In the third step, we construct quarterly route level and airport level data from the T-100
and DB1B databases.13 In calculating average non-stop fares, zero fares and abnormally high
fares were excluded from the dataset. We only use average fares which are based on at least
ten observations and thousand quarterly passengers. We add demographic information on the
labor force, average income, and the number of establishments of the respective Metropolitan
Statistical Areas from the U.S. Bureau of Labor Statistics. Applying this procedure, we arrive
at a quarterly panel dataset of 1,258 non-stop routes allowing a detailed econometric
investigation of the effects of firm exit.
4.2 EMPIRICAL APPROACH
Guided by our conceptual framework derived in Section 2 above, our empirical approach can
be subdivided into five consecutive steps. In the first step, we use fixed effects regression
models to estimate the short run effects of firm exit events on average market yield,
departures and passengers.14 In the second step, we extend the observation window after exit
and rerun the respective regressions for medium- and long-term in order to investigate
possible changes, e.g., due to the realization of merger efficiencies. The third step introduces
an interaction term allowing the effects of exit to differ on routes which became a monopoly
post-exit. Subsequently, in the fourth step, we refrain from holding the number of carriers
constant after an exit-event and examine the entry-inducing effects of firm exits. In the fifth
and last step, we exclude the three large mergers (Delta Air Lines – Northwest Airlines, US
Airways - America West Airlines and American Airlines – Trans World Airlines) from the
analysis and rerun all regressions. This step allows us to investigate whether our results hold
for the sub-sample of smaller mergers.
Discussing the technicalities of the five steps of our empirical approach in greater detail,
our variables of interest are the exit variables which are captured by four dummy variables.
We distinguish between two types of merger-related exits. In the first case, only the exiting
carrier was active on the respective non-stop route and the resulting entity inherited this route.
We call this a ‘route switching’ merger exit. In that case, the exit does not trigger a change in
the number of carriers. If both of the two merging parties have provided non-stop service
before, we call this an ‘overlapping route’ merger exit. This type differs from a merger exit
with route switching since the number of competing carriers is reduced by one carrier. Exits
13 At this step, operations of regional carriers are merged with the operations of their respective network carrier. If a regional carrier operates flights for more than one network carrier, the network carrier is assigned on a route-by-route basis according to the hub airport involved. This procedure was cross-checked with information on the ticketing carrier on the respective routes provided by the DB1B database.
14 We do not report the results for ‘seats’ as they are very closely related to ‘departures’.
15
which followed the liquidation of an airline are called liquidation exits. Operational exits are
all other route exits which are not directly related to a firm exit. This category includes, e.g.,
network restructuring exits or exits due to unprofitability. Exits which were either observed in
times of financial distress15 or took place before the merger was settled are also classified as
operational exits.
Turning to our estimation approach, we first estimate several log-linear fixed effects
regression models which can be denoted by
0
4
2
ln( )
,
it opEx it mExSw it mExOv it lqEx it
X it year t qj jt i itj
y opEx mExSw mExOv lqEx
X year quarter
(1)
where yit is either the non-stop yield (fare per passenger mile), the number of departures or the
number of passengers transported. The variable opEx captures operational exits, mExSw
captures switching merger exits, mExOv captures merger exits on overlapping routes, and
lqEx captures exits due to the liquidation of a carrier. The different exit dummies are zero
before the exit event and become one in the quarter after exit and the subsequent quarter(s)
depending on whether short-, medium-, or long-term effects shall be assessed. To capture the
short-term effects of entry we compare non-stop fares, departures, and passengers eight
quarters before exit with the first two quarters after exit. Thus, the exit variable is one for two
quarters. Two quarters after exit the observation periods ends. Respectively, we capture
medium-term effects by following prices and quantity four quarters after exit and long-term
effects by following prices and quantities eight quarters after exit16. The introduction of such
a ‘dynamic’ perspective allows us to investigate whether the observed short-term effects are
permanent or rather disappear due to, e.g., realized merger efficiencies or competitive
reactions by competitors. Thus, the coefficient estimates of the different exit variables report
the average percentage change in prices and quantity after a certain type of exit. We further
include a set of route-, airport- or MSA-specific control variables (X) as well as a yearly trend
(year) and seasonal dummies (quarter).
15 Since it is unclear whether financial distress, e.g., filing for Chapter 11, ends in the liquidation of a firm, exits which occur before the liquidation of an airline are interpreted as operational exits since they basically aim at restoring profitability. As a consequence, the group of operational exits is quite heterogeneous. However, as we aim at using this category for the purpose of comparison only, we refrain from a further differentiation.
16 Since the dataset covers the period up to the third quarter of 2011, the long-term effects of the Delta-Northwest merger (and all other exits which took place either in the fourth quarter of 2009 or in the first quarter of 2010) refer to the first seven quarters after exit.
16
As control variables, we include the number of carriers without the exiting carrier or
merging parties (# airlines w/o exit) and the number of low-cost carriers, also without the
exiting carrier or merging carriers (# LCCs w/o exit). These variables account for the effect of
market structure over time. We further control for the average size of planes the carriers use to
serve the route (avg. plane size) since costs should decline with an increasing capacity of the
aircraft. When estimating the price effects of exit, we also include the average one-stop yield
(ln(one-stop yield)) to account for possible price competition from connecting flights.17 We
also control for the influence of airport size as measured by the mean of the two endpoint
airports’ passenger share (airport size (mean)). Furthermore, three demographic variables on
the MSA level enter the analysis which aim to capture demand effects. The labor force
(ln(labor force) (mean)) shall capture potential total demand. The number of establishments
(ln(# establ.) (mean)) is included to capture the demand of less price-sensitive business people
and regional economic prosperity shall be captured by the average weekly wage in the
respective MSAs (ln(avg. weekly wage) (mean)).
As it is reasonable to assume that the size of the competitive effects under investigation
depends on the post-exit market structure, in a next step, we introduce an interaction term
which allows isolating the effects of exits on routes which resulted in a monopoly post-exit.
The fixed effects regression model becomes
0
4
2
ln( )
.
it opEx it mExSw it mExOv it lqEx it
opExM it i mExSwM it i
mExOvM it i lqExM it i
X it year t qj jt i itj
y opEx mExSw mExOv lqEx
opEx mono mExSw mono
mExOv mono lqEx mono
X year quarter
(2)
In this model approach, the coefficients of the exit dummies alone denote the average
percentage change in prices or quantities, respectively, if there are at least two competitors
left directly after the exit event. If the market structure turns from a duopoly to a monopoly
after the exit18, the effects of exit can be calculated as the sum of the respective coefficients
(e.g., βopEx+ βopExM).
17 The one-stop yield is missing if either the route is not served via connecting flights or if there are not enough observations in DB1B data to be able to calculate a reliable mean (see section 4.1). In order to avoid losing a substantial amount of observations for regression analysis, an arbitrary value is assigned to these observations and an additional dummy variable is included which marks these observations (missing one-stop yield). This method is called dummy variable adjustment or missing indicator method and is frequently used in econometric analysis (Allison, 2001).
18 The route might also stay a monopoly in case of switching route (merger) exits.
17
For an assessment of possible entry-inducing effects of firm exit, in the fourth step, we
estimate a similar model as specified in equation (1) above. The dependent variable becomes
the change in the number of carriers other than the exiting one or the merging parties ( #
airlines w/o exit). Accordingly, we refrain from holding the number of other carriers constant
but include the lagged value of this variable since the previous competitive environment
should largely determine entry activity of other carriers after exit events.
In the fifth and last step, we exclude the three large mergers (Delta Air Lines – Northwest
Airlines, US Airways - America West Airlines and American Airlines – Trans World
Airlines) from the analysis and rerun the regressions. This step allows us to investigate
whether our results hold for the sub-sample of smaller mergers.
4.3 DESCRIPTIVE STATISTICS
As already mentioned in Section 4.1, the dataset covers 1,258 route exits. The majority of
these exits are operational exits (918 exits). We further observe 217 merger exits on switching
routes and 79 merger exits on overlapping routes. 44 exits occurred because of carrier
liquidations. Directly after exit we observe that about 40 percent of the routes are monopolies
(see Table 3).
Table 3: Route exits included in the fixed effects regressions
# of exits share of post-exit monopolies
operational exits 918 42.48% merger exits (switching) 217 27.65% merger exits (overlap) 79 46.84% liquidation exits 44 31.82% Total 1,258 39.83%
Sources: U.S. DOT, T-100 Domestic Segment Data, authors’ calculations.
About 28% of the switching route exits are monopolies. In contrast to the other exit types,
these routes have also been monopolies before the exit event as no change in the number of
carriers was triggered by the exit event. Further summary statistics for the variables included
in the regressions can be retrieved from Table 4.
18
Table 4: Summary statistics
Quarter before/of exit Quarter after exit Period before exit Period after exitVariable mean s.d. mean s.d. mean s.d. mean s.d. ln(non-stop yield) 2.787 (0.652) 2.818 (0.651) 2.823 (0.658) 2.825 (0.650) non-stop yield 20.448 (16.323) 21.093 (16.684) 21.301 (16.994) 21.289 (17.089) ln(departures) 7.008 (0.809) 6.922 (0.877) 7.005 (0.826) 6.960 (0.860) Departures 1,497 (1,241) 1,425 (1,233) 1506 (1272) 1469 (1275) ln(Passengers) 11.416 (0.898) 11.306 (0.970) 11.434 (0.899) 11.361 (0.950) Passengers 131,621 (119,584) 123,061 (116,915) 133,567 (120637) 128,696 (121,496) # airlines w/o exit 0.045 (0.332) 0.217 (0.485) 0.012 (0.294) 0.028 (0.329) operational exit - - 0.730 (0.444) - - 0.746 (0.436) merger exit (switching) - - 0.172 (0.378) - - 0.166 (0.372) merger exit (overlap) - - 0.063 (0.243) - - 0.052 (0.222) liquidation exit - - 0.035 (0.184) - - 0.036 (0.187) post-exit monopoly 0.398 (0.490) 0.398 (0.490) 0.398 (0.490) 0.393 (0.489) # airlines w/o exit 1.417 (1.000) 1.634 (1.013) 1.354 (1.005) 1.662 (0.977) # LCCs w/o exit 0.401 (0.551) 0.412 (0.563) 0.356 (0.526) 0.445 (0.592) avg. plane size 122.185 (41.045) 119.790 (43.943) 126.164 (40.056) 120.170 (44.180) ln(one-stop yield) 2.711 (1.077) 2.778 (0.950) 2.799 (0.960) 2.782 (0.970) missing one-stop yield 0.021 (0.142) 0.013 (0.112) 0.013 (0.113) 0.014 (0.116) airport size 1.890 (0.797) 1.881 (0.794) 1.899 (0.786) 1.886 (0.800) ln(# establ. ) 11.557 (0.714) 11.559 (0.715) 11.541 (0.714) 11.578 (0.718) ln(avg. weekly wage) 6.696 (0.174) 6.702 (0.167) 6.666 (0.175) 6.721 (0.164) ln(labor force) 14.465 (0.654) 14.467 (0.654) 14.455 (0.655) 14.478 (0.655) Year 2004 (3.802) 2004 (3.898) 2003 (3.880) 2005 (3.817) Quarter 2 0.208 (0.406) 0.183 (0.387) 0.250 (0.433) 0.258 (0.437) Quarter 3 0.297 (0.457) 0.208 (0.406) 0.250 (0.433) 0.240 (0.427) Quarter 4 0.312 (0.464) 0.297 (0.457) 0.249 (0.432) 0.244 (0.430) Observations 1,258 1,258 10,037 9,752 Notes: Prices in 1995 $ cents. Sources: U.S. DOT, T-100 Domestic Segment Data, Airline Origin and Destination and U.S. Bureau of Labor Statistics, authors’ calculations.
Our first dependent variable is the non-stop yield which is measured in real 1995 U.S.
cents per passenger mile. As shown in Table 4, in the quarter before/of exit, a passenger paid
about 20.4 cents per mile. The yield has risen to 21.1 cents per mile on average in the quarter
after exit. Comparing the average price over the entire period before the exit (7-8 quarters)
with the average price over the entire period after exit, we do not observe a price increase in
real terms. The second dependent variable is the number of quarterly departures which
amounts to 1,497 departures on average in the quarter before/of exit. It drops to 1,425
departures in the quarter after exit. Although the extension of the observation period leads to a
moderate increases in the number of departures, the pre-exit level remains unreached. The
third dependent variable is the number of quarterly passengers. On average, about 132,000
passengers have been transported in the quarter before/of exit. Over the quarter after exit the
number of passengers drops to about 123,000 passengers. Although an extension of the
analysis to the entire observation period after exit shows an increase in the number of
19
passengers, this increase again turns out to be insufficient to restore pre-exit levels. For the
regressions which aim to assess the entry-inducing effect of exit, the fourth dependent
variable is the change in the number of other carriers. While in the quarter before/of exit, this
number amounts to 0.045 exits on average it increases to 0.217 exits in the quarter after exit.
Both values experience a substantial drop if the analysis is extended to the entire period
before/after exit.
Descriptive statistics distinguished by the type of exit are provided in Table 9 to Table 12
in the Appendix. From these bivariate statistics the effect of exit seems to be more
pronounced for operational and liquidation exits than for merger-related exits; this is true for
price, quantity and entry-inducing effects. Interestingly, although bankrupt carriers were
much smaller than most of the merging carriers, routes which are subject to the different exits
do not differ substantially in size. Non-stop routes which were subject to a liquidation exit
transported about 190,000 passengers in the quarter before/of exit, while routes subject to
merger exits have been travelled less (switching: 161,000 passengers; overlapping: 129,000
passengers) in the quarter before/of exit.
4.4 EMPIRICAL RESULTS AND INTERPRETATION
Based on the description of our dataset and the empirical approach, this section presents our
empirical results and interpretation. We subdivide our discussion into the reporting of the key
empirical results for firm exits through liquidation and through merger. Results of the
regressions with route fixed effects on non-stop yield can be retrieved from Table 5. Table 6
depicts the results on the number of departures, and Table 7 presents the regression results
with the number of passengers as our dependent variable. Each table is split into three panels.
The first panel shows the short-term regressions, the second panel shows the medium-term
regressions, and the third panel shows the long-term regressions. Within each panel, the first
column does not include the post-exit monopoly interaction term while the second column
does. The effects for operational exits are included for the purpose of comparison.19
4.4.1 FIRM EXIT THROUGH LIQUIDATION
In the short run, exit through liquidation is found to have a substantial effect on market yield.
On average, prices increase by 12 percent in the first two quarters after exit. When allowing
for differences regarding the post-exit market structure, the effect does not differ significantly
19 An average operational exit is found to have a similar ceteris paribus effect on non-stop yield in the short-, medium-, and long-term. After an operational exit, non-stop yield increase by 6 percent on average, the number of departures decrease by 19 to 20 percent, and passengers transported drop by about 17 percent.
20
between post-exit monopolies and routes with at least two competitors. Interestingly, the
effect is found to be persistent over time. The substantial size of the effects of liquidation
exits on price and quantity measures can be substantiated by a direct comparison to the results
for the ‘operational exit’ group: in the short run, we find an average yield increase of about 6
percent which turns out to be quite robust for a narrower focus on monopoly routes post-exit,
extensions of the observation window and the sub-sample containing only small mergers.
Turning from price effects to quantity effects, it is found that firm exits due to liquidation
lead to the expected large decrease in capacity and demand. In the short run, liquidation exits
cause an average decrease in the number of departures of 17.4 percent and an average
decrease in demand of 14.8 percent. The effect for the number of passengers is significantly
and substantially higher if only monopoly routes post-exit are taken into account (-28.4
percent) compared to an effect of -7.5 percent for routes with more than one competitor.
Again, the results appear to be quite robust for extensions of the observation window and the
sub-sample containing only smaller mergers. Interestingly, while price effects of exits due to
firm liquidations were significantly higher than the price effects of operational exits, the
quantity effects do not significantly differ between liquidation exits and operational exits.
Assessing the effect of liquidation exits on market entry, we find a substantial effect
already in the short-run (see Table 8). The change in the number of carrier increases by 0.130
after a liquidation exit and controlling for other factors. This effect is smaller than for
operational exits (0.192), but the two effects do not differ significantly. Again, the effect is
persistent over time. In general, since market entry is expected to create a downward pressure
on price (see, e.g., Hüschelrath and Müller, 2011, or Daraban and Fournier, 2009), the entry
activity of other carriers induced by multiple market exits of liquidated carriers should act as a
countervailing force regarding price increases and quantity reductions. However, consulting
our descriptive results reported in Table 12 in the Appendix, it becomes apparent that the
induced entry activity is not sufficient to fully compensate the price increase observed
immediately after a liquidation exit. Even adjusted for inflation, we observe the yield to be
about 1 cent per mile (in 1995$ terms) higher in the period after exit than in the period before
exit (see Table 12).
In a nutshell, our empirical findings imply that, even if the total share of the failed firms’
domestic passengers was negligible, these firms’ market exits had a significant impact on the
respective routes. These markets suffered, ceteris paribus, from substantial price increases and
service reductions (in the form of lower flight frequencies). However, incentives for entry
seem to be high since there is a substantial increase in the change of the number of other
21
carriers immediately after the observed liquidation exits. This finding suggests that liquidation
exits may (at least partly) cause a welfare-improving replacement of the inefficient bankrupt
carrier with a more efficient operating airline. However, this comes at a cost for consumers as
entries after liquidation exits cannot fully reverse the price increase observed immediately
after exit.
Table 5: Fixed effects regressions for the effect of exits on non-stop yield
ln(non-stop yield) - short-term ln(non-stop yield) - medium term ln(non-stop yield) - long termVariable coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) operational exit 0.060 *** (0.005) 0.055*** (0.006) 0.067*** (0.005) 0.059*** (0.006) 0.061*** (0.005) 0.054*** (0.006) merger exit (switching) 0.027 *** (0.008) 0.036*** (0.009) 0.027*** (0.009) 0.040*** (0.009) 0.010 (0.010) 0.030*** (0.010) merger exit (overlap) 0.056 *** (0.015) 0.025 (0.024) 0.047*** (0.016) 0.029 (0.023) 0.027 (0.018) 0.022 (0.024) liquidation exit 0.120 *** (0.017) 0.106*** (0.020) 0.127*** (0.016) 0.114*** (0.018) 0.117*** (0.017) 0.109*** (0.019) op. exit # monopoly 0.013 (0.009) 0.018** (0.009) 0.016* (0.009) m. ex. (sw.) # monopoly -0.034** (0.017) -0.047** (0.019) -0.071*** (0.022) m. ex. (ov.) # monopoly 0.068** (0.029) 0.037 (0.030) 0.009 (0.035) liq. exit # monopoly 0.042 (0.035) 0.040 (0.036) 0.025 (0.036) # airlines w/o exit -0.021 *** (0.006) -0.020*** (0.006) -0.023*** (0.005) -0.021*** (0.005) -0.035*** (0.005) -0.034*** (0.005) # LCCs w/o exit -0.101 *** (0.014) -0.100*** (0.014) -0.103*** (0.012) -0.102*** (0.012) -0.093*** (0.011) -0.092*** (0.011) avg. plane size -0.001 *** (0.000) -0.001*** (0.000) -0.001*** (0.000) -0.001*** (0.000) -0.000*** (0.000) -0.000*** (0.000) ln(one-stop yield) 0.377 *** (0.025) 0.376*** (0.025) 0.404*** (0.023) 0.403*** (0.023) 0.387*** (0.029) 0.386*** (0.029) missing one-stop yield 2.600 *** (0.176) 2.592*** (0.177) 2.798*** (0.166) 2.791*** (0.166) 2.672*** (0.206) 2.663*** (0.205) airport size (mean) -0.033 * (0.019) -0.034* (0.019) -0.018 (0.018) -0.021 (0.018) 0.013 (0.019) 0.008 (0.019) ln(# establ. ) (mean) 0.463 *** (0.092) 0.461*** (0.093) 0.421*** (0.084) 0.414*** (0.086) 0.247*** (0.075) 0.232*** (0.076) ln(avg. weekly wage) 0.228 *** (0.050) 0.227*** (0.050) 0.216*** (0.046) 0.211*** (0.046) 0.254*** (0.048) 0.245*** (0.047) ln(labor force) (mean) 0.579 *** (0.188) 0.584*** (0.189) 0.528*** (0.183) 0.514*** (0.184) 0.490*** (0.167) 0.459*** (0.167) Year -0.055 *** (0.004) -0.055*** (0.004) -0.048*** (0.004) -0.048*** (0.004) -0.031*** (0.003) -0.031*** (0.003) Quarter 2 -0.014 *** (0.003) -0.015*** (0.003) -0.013*** (0.003) -0.013*** (0.003) -0.006** (0.003) -0.007** (0.003) Quarter 3 -0.039 *** (0.004) -0.038*** (0.004) -0.031*** (0.004) -0.031*** (0.004) -0.023*** (0.003) -0.024*** (0.003) Quarter 4 -0.069 *** (0.005) -0.069*** (0.005) -0.061*** (0.005) -0.060*** (0.005) -0.052*** (0.004) -0.052*** (0.004) Constant 96.151 *** (7.352) 96.335*** (7.394) 84.022*** (6.470) 83.689*** (6.498) 52.762*** (5.093) 52.069*** (5.081) R2 (within/between/overall) 0.283/0.109/0.110 0.285/0.108/0.109 0.287/0.141/0.142 0.289/0.147/0.147 0.267/0.199/0.200 0.270/0.222/0.223 Observations 12,553 12,553 15,069 15,069 19,789 19,789 Routes 1,258 1,258 1,258 1,258 1,258 1,258
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data, Airline Origin and Destination Survey (DB1B) and U.S. Bureau of Labor Statistics, authors’ calculations.
Table 6: Fixed effects regressions for the effect of exits on number of departures
ln(departures) - short term ln(departures) - medium term ln(departures) - long termVariable coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) operational exit -0.200 *** (0.011) -0.123*** (0.011) -0.192*** (0.011) -0.121*** (0.011) -0.189*** (0.011) -0.129*** (0.012) merger exit (switching) -0.021 * (0.012) -0.023 (0.014) -0.026** (0.012) -0.036*** (0.013) -0.035*** (0.013) -0.045*** (0.015) merger exit (overlap) -0.031 * (0.017) -0.055*** (0.019) -0.033* (0.017) -0.059*** (0.021) -0.046** (0.018) -0.076*** (0.023) liquidation exit -0.174 *** (0.032) -0.146*** (0.036) -0.175*** (0.029) -0.149*** (0.034) -0.180*** (0.030) -0.165*** (0.038) Op. exit # monopoly -0.170*** (0.024) -0.158*** (0.022) -0.137*** (0.021) m. ex. (sw.) # monopoly 0.002 (0.027) 0.033 (0.027) 0.041 (0.028) m. ex. (ov.) # monopoly 0.048 (0.032) 0.051 (0.032) 0.064* (0.035) liq. exit # monopoly -0.079 (0.070) -0.073 (0.054) -0.039 (0.053) # airlines w/o exit 0.151 *** (0.014) 0.137*** (0.014) 0.149*** (0.014) 0.137*** (0.014) 0.142*** (0.014) 0.137*** (0.014) # LCCs w/o exit 0.089 *** (0.025) 0.088*** (0.025) 0.106*** (0.023) 0.103*** (0.023) 0.106*** (0.020) 0.101*** (0.020) Avg. plane size -0.003 *** (0.001) -0.004*** (0.001) -0.004*** (0.001) -0.004*** (0.001) -0.004*** (0.001) -0.004*** (0.000) airport size (mean) 0.773 *** (0.044) 0.776*** (0.043) 0.780*** (0.045) 0.786*** (0.043) 0.788*** (0.043) 0.791*** (0.041) ln(# establ. ) (mean) 0.201 (0.170) 0.123 (0.168) 0.332** (0.166) 0.236 (0.162) 0.368** (0.152) 0.286* (0.149) ln(avg. weekly wage) (mean) 0.527 *** (0.093) 0.530*** (0.092) 0.594*** (0.089) 0.606*** (0.088) 0.587*** (0.087) 0.605*** (0.086) ln(labor force) (mean) 0.443 (0.359) 0.452 (0.358) 0.234 (0.350) 0.285 (0.351) 0.001 (0.319) 0.068 (0.319) Year -0.005 (0.008) -0.003 (0.007) -0.005 (0.007) -0.004 (0.007) -0.004 (0.006) -0.003 (0.006) Quarter 2 0.061 *** (0.006) 0.062*** (0.006) 0.060*** (0.006) 0.061*** (0.006) 0.062*** (0.006) 0.063*** (0.006) Quarter 3 0.061 *** (0.009) 0.062*** (0.009) 0.062*** (0.008) 0.063*** (0.008) 0.060*** (0.007) 0.062*** (0.007) Quarter 4 -0.003 (0.008) -0.001 (0.008) -0.008 (0.008) -0.007 (0.008) -0.008 (0.007) -0.007 (0.007) Constant 2.945 (13.034) 0.935 (12.989) 4.803 (12.092) 3.239 (11.992) 5.197 (9.959) 4.357 (9.889) R2 (within/between/overall) 0.216/0.296/0.285 0.230/0.303/0.293 0.232/0.310/0.298 0.248/0.318/0.307 0.249/0.339/0.326 0.262/0.347/0.334 Observations 12,553 12,553 15,069 15,069 19,789 19,789 Routes 1,258 1,258 1,258 1,258 1,258 1,258
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data and U.S. Bureau of Labor Statistics, authors’ calculations.
Table 7: Fixed effects regressions for the effect of exits on number of passengers
ln(passengers) – short term ln(passengers) – medium term ln(passengers) – long termVariable coef. (s.e). coef. (s.e) coef. (s.e) coef. (s.e) coef. (s.e) coef. (s.e) operational exit -0.172 *** (0.011) -0.097*** (0.011) -0.167*** (0.011) -0.095*** (0.011) -0.170*** (0.011) -0.108*** (0.011) merger exit (switching) -0.042 *** (0.012) -0.040*** (0.013) -0.044*** (0.012) -0.055*** (0.013) -0.051*** (0.013) -0.066*** (0.014) merger exit (overlap) -0.053 *** (0.016) -0.059*** (0.019) -0.043** (0.017) -0.051** (0.022) -0.066*** (0.019) -0.080*** (0.024) liquidation exit -0.148 *** (0.040) -0.075** (0.037) -0.159*** (0.039) -0.091** (0.037) -0.179*** (0.036) -0.129*** (0.040) Op. exit # monopoly -0.166*** (0.024) -0.160*** (0.022) -0.143*** (0.021) m. ex. (sw.) # monopoly -0.010 (0.026) 0.040 (0.026) 0.057** (0.027) m. ex. (ov.) # monopoly 0.008 (0.031) 0.014 (0.033) 0.030 (0.036) liq. exit # monopoly -0.219** (0.091) -0.205** (0.084) -0.149** (0.069) # airlines w/o exit 0.135 *** (0.013) 0.122*** (0.013) 0.133*** (0.013) 0.121*** (0.013) 0.129*** (0.013) 0.124*** (0.012) # LCCs w/o exit 0.115 *** (0.026) 0.113*** (0.026) 0.133*** (0.023) 0.128*** (0.023) 0.133*** (0.021) 0.126*** (0.021) Avg. plane size 0.005 *** (0.001) 0.005*** (0.001) 0.005*** (0.001) 0.005*** (0.001) 0.004*** (0.001) 0.004*** (0.001) airport size (mean) 1.051 *** (0.047) 1.052*** (0.046) 1.032*** (0.048) 1.036*** (0.046) 0.996*** (0.045) 0.999*** (0.044) ln(# establ. ) (mean) 0.192 (0.168) 0.121 (0.163) 0.385** (0.163) 0.298* (0.157) 0.477*** (0.150) 0.400*** (0.147) ln(avg. weekly wage) (mean) 0.367 *** (0.096) 0.366*** (0.095) 0.475*** (0.092) 0.484*** (0.091) 0.513*** (0.088) 0.529*** (0.088) ln(labor force) (mean) 0.725 ** (0.359) 0.709** (0.357) 0.431 (0.347) 0.179 (0.314) Year 0.004 (0.008) 0.006 (0.008) 0.003 (0.007) 0.004 (0.007) 0.006 (0.006) 0.007 (0.006) Quarter 2 0.143 *** (0.007) 0.144*** (0.007) 0.143*** (0.007) 0.144*** (0.007) 0.147*** (0.006) 0.148*** (0.006) Quarter 3 0.130 *** (0.009) 0.130*** (0.009) 0.132*** (0.008) 0.133*** (0.008) 0.133*** (0.008) 0.134*** (0.008) Quarter 4 0.035 *** (0.008) 0.036*** (0.008) 0.026*** (0.008) 0.028*** (0.008) 0.028*** (0.007) 0.029*** (0.007) Constant -14.581 (13.345) -17.182 (13.301) -10.094 (12.454) -12.047 (12.378) -14.374 (10.294) -15.403 (10.280) R2 (within/between/overall) 0.296/0.352/0.340 0.308/0.361/0.350 0.296/0.375/0.361 0.311/0.382/0.369 0.294/0.405/0.388 0.307/0.410/0.394 Observations 12,553 12,553 15,069 15,069 19,789 19,789 Routes 1,258 1,258 1,258 1,258 1,258 1,258
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data and U.S. Bureau of Labor Statistics, authors’ calculations.
Table 8: Fixed effects regressions for the effects of exits on entry
# airlines w/o exiting/merger – short term
# airlines w/o exiting/merger – medium term
# airlines w/o exiting/merger - long term
Variable coef. (s.e.) coef. (s.e.) coef. (s.e.) operational exit 0.192*** (0.014) 0.152*** (0.012) 0.141 *** (0.011) merger exit (switching) -0.017 (0.014) 0.011 (0.012) 0.056 *** (0.011) merger exit (overlap) 0.003 (0.024) 0.012 (0.018) 0.049 *** (0.016) liquidation exit 0.130*** (0.042) 0.107** (0.044) 0.137 *** (0.040) # airlines w/o exit (lag) -0.392*** (0.017) -0.350*** (0.014) -0.303 *** (0.012) # LCCs w/o exit (lag) -0.045 (0.029) -0.028 (0.025) -0.021 (0.020) Avg. plane size -0.003*** (0.000) -0.003*** (0.000) -0.003 *** (0.000) airport size (mean) 0.325*** (0.042) 0.297*** (0.037) 0.268 *** (0.029) ln(# establ. ) (mean) 0.047 (0.236) 0.069 (0.194) 0.049 (0.126) ln(avg. weekly wage) (mean) 0.161 (0.117) 0.020 (0.097) -0.010 (0.080) ln(labor force) -1.633*** (0.389) -1.273*** (0.329) -0.923 *** (0.237) Year 0.025*** (0.008) 0.009 (0.007) -0.009 ** (0.005) Quarter 2 0.022** (0.009) 0.013 (0.008) 0.015 ** (0.007) Quarter 3 -0.017* (0.009) -0.028*** (0.008) -0.026 *** (0.007) Quarter 4 0.018* (0.011) 0.015 (0.010) 0.008 (0.008) Constant -27.270** (13.365) -0.367 (12.281) 31.832 *** (8.924) R2 (within/between/overall) 0.214 / 0.001 / 0.004 0.191 / 0.000 / 0.005 0.170 / 0.001 / 0.007 Observations 12,512 15,028 19,748 Routes 1,258 1,258 1,258
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data and U.S. Bureau of Labor Statistics, authors’ calculations.
4.4.2 FIRM EXIT THROUGH MERGER
Following our conceptual framework derived in Section 2, the analysis of the effects of firm
exits through merger must differentiate between two route types: overlapping routes and
switching routes. In the short run, exit through merger has a significant effect on both route
types: while prices increase by on average about 5.6 percent on overlapping routes, switching
routes experience a yield increase of about 2.7 percent. The extension of the observation
window, however, reveals that these effects vanish in the long run. While effects are still
significantly different from zero in the medium term, prices are not significantly higher for
both types of merger exits in the long term.
Interestingly, if the focus is narrowed down to monopoly routes post-exit, the effect on
switching routes is found to be zero (as the coefficient of the interaction term is negatively
significant and of approximately the same absolute size) in the short run and even negative (-
4.1 percent) in the long run. While the above effect on overlapping routes is in the short run
mostly driven by post-exit monopolies, results diverge in the medium and long run. The
extension of the observation window reveals that the effect is of only half the size and no
26
longer significant. For the sub-sample of small mergers (see Table 13), we do not find any
significant effects of merger exits on average yield. While one should refrain from
interpreting the results for overlapping routes due to the low number of observations in this
category20, the insignificant effect found on switching routes is not surprising since the
possible drivers for the positive effect in case of larger mergers, i.e., reduced competition due
to multimarket contact and the elimination of potential entrants, might be negligible for small
firm acquisitions.
Turning from the price effects to the quantity effects (see Table 6 and Table 7), it is found
that firm exits through merger lead to smaller reductions in capacity and demand than
liquidation-related and operational exits. In the short run, the number of departures is reduced
by about 3.1 percent on overlapping routes and 2.1 percent on switching routes. In markets
which turn to a monopoly post-exit, we fail to find any significant effect (for both route types)
on the number of departures since the sum of the coefficients is not significantly different
from zero (as indicated by the Wald test). Demand declines in both cases, however, the effects
do not significantly differ between post-exit monopolies and post-exit oligopolies.
Furthermore, these demand decreases in post-exit monopolies are only found in the short term
as the sum of the merger exit coefficients and the coefficients of the respective interaction
terms are not significantly different from zero in the medium and long run. In contrast, the
analysis of the non-monopoly routes provides a different picture. The short-term reductions in
the number of departures and the number of passengers are found to be even more
pronounced in the medium and long run: the reduction in the number of departures increase
from about 5.5 percent (short-term) to 7.6 percent (long-term) for the overlapping routes and
from about 2.3 percent (short-term, insignificant) to 4.5 percent (long-term, highly significant)
for the switching routes. For the sub-sample of smaller mergers (see Table 14 and Table 15),
the quantity effects found for the switching routes in non-monopoly markets are even more
pronounced than for the whole sample showing reductions in the number of departures from
7.2 percent in the short-term to 9.3 percent in the long-term and reductions in the number of
passengers from 7.3 percent in the short-term to 10 percent in the long-term.
Regarding potential entry-inducing effects, we find that the change in the number of
carriers does not react after merger-related exits in the short- and medium-term, but increases
20 After excluding the exits of the six carriers involved in the large mergers, 38 routes for merger exits on switching routes, 9 routes for merger exits on overlapping routes and 799 routes for operational exits remain in the dataset.
27
by 0.056 carriers (switching routes) and 0.049 carriers (overlapping routes) in the long-term.
On the contrary, for the sub-sample of small-mergers (see Table 16) we find that entry
materializes even in the medium and short run. Furthermore, the entry-inducing effects of
small mergers are substantially larger than for the group of all mergers and of similar size as
the entry-inducing effects of liquidation exits and operational exits. As we see from the
descriptive statistics both for the merger exits with route switching (Table 10) and the merger
exits on overlapping routes (Table 11), the countervailing effect of induced entries seems to
be sufficient to drive down real prices (at least in the long term) to pre-exit levels.
In a nutshell, our analysis of the effects of merger-related firm exits shows that price
increases and quantity reductions are not only an issue on the overlapping parts of the
merging firm’s networks but can also play a role on its complementary parts (so-called
switching routes). These results support previous findings on reduced competition due to an
increase in multimarket contact, however, might also be explained by changes in the pricing
strategy of the acquired carrier, increased airport dominance, or the increase in quality
achieved by the interconnection of the two networks. Interestingly, in case of monopoly
routes (in which prices have already been high before the merger was settled) we do not
observe any price increases post-merger but even a price decrease in the long run. Although
this finding might not be too surprising – basically because the merged entity would reduce
profits by raising the price above the existing monopoly level – the observed yield reductions
can be interpreted as an indication for the realization of merger efficiencies (which also a
monopolist is partly passing-on downstream to the final customers in the form of price
reductions).
The observed effects on overlapping routes are largely found to be driven by the large
mergers (especially the Delta Air Lines – Northwest Airlines merger) since networks have
been (almost) fully complementary for the smaller mergers. Although quantity effects have
been persistent and concentration has risen substantially, the price effects are mostly found in
the short run and vanish over time. This can also be interpreted as a further indication for
realized merger efficiencies.
Entry is found to take place rather quickly in case of liquidation exits and small mergers.
However, for the entire group of mergers entry-inducing effects only materialize in the long
run. It can therefore be concluded that particularly large mergers appear to (temporarily)
reduce the incentives of other carriers to enter the respective routes. This observation is
28
reasonable as the merger led to increased market shares of the merged entity thereby, ceteris
paribus, reducing the entry incentives of competitors. However, in the long run, entry might
become attractive again, e.g., due to restructuring activities of the merged carrier and/or
quality problems that materialize during the post-merger integration process.
4.5 POLICY IMPLICATIONS
Our empirical results allow the derivation of several important policy conclusions especially
for unilateral effects analysis as part of horizontal merger assessments. First, given our result
that the price and quantity effects of liquidation exits are much more pronounced than the
respective effects of merger-related exits, on the surface, it could be concluded that avoiding
liquidation through merger benefits consumers and society. In other words, our results
apparently support the so-called failing firm defense which allows the clearance of (partly)
anticompetitive mergers in cases in which one of the merging firms is at the verge of
bankruptcy. However, despite the obvious advantages of smaller price increases and better
service options, the net welfare effects of such ‘failing firm mergers’ remain unclear, e.g., due
to possible negative effects on the merged carrier and the industry triggered by the prevention
of the exit of (inefficient) capacity from the industry (and dissuading entry by potential
competitors) or substantial problems in the integration of the merging carriers. Furthermore, it
must be reminded that our data set only contains mergers which have been approved by the
Department of Justice (DOJ). Although wrong decisions on the side of the DOJ cannot be
ruled out completely, it seems very unlikely that the approved mergers were in fact
anticompetitive and should have been prohibited.
Second, with respect to mergers, our results reveal that larger mergers, ceteris paribus, lead
to significant and partly permanent price increases first and foremost not – as expected by
theory – on the overlapping parts of the route network but on routes in which the operating
airline simply switched as a consequence of the merger. Any analyst studying the competitive
effects of a horizontal merger is therefore well advised to consider potential effects not only
on the overlapping part of the network but on non-overlapping parts as well in order to come
to meaningful conclusions on the unilateral effects of the merger proposal.
Third, still focusing on mergers, we find evidence for the realization of merger efficiencies
for large mergers only. While in the short run, significant price increases on both switching
and overlapping routes were observed, these effects either disappear or are reduced
substantially in the long run. As we control for the number of firms (and the other key yield
29
drivers) in our regressions, the observed price reductions must be associated with the
realization and the pass-on of merger efficiencies. Furthermore, our finding of significant
price reductions for the case of monopoly routes – in which by definition competition is
excluded as alternative driver of price reductions – can be interpreted as further indication for
the existence of merger efficiencies which are (at least partly) passed on to the customers in
the form of lower prices. Although antitrust policy might still be well-advised to keep up the
rather skeptical approach with respect to merger efficiencies when it comes to a weighting of
pro- and anticompetitive effects as part of the merger control procedure, our study suggests
that these efficiencies are existent to a degree that allows the reversion of the price increases
observed immediately after the completion of the merger.
Fourth, our analysis shows that especially large mergers have an entry-dissuading effect in
the short and medium run. In contrast, liquidation-related exits are found to cause immediate
entry-inducing effects. Although an econometric analysis of the price effects of these
subsequent entries is beyond the scope of this paper, theory and descriptive evidence
presented above suggests for the group of merger-related exits that they are strong enough to
reverse the short-term price increases post-merger.21 Although this finding is an encouraging
sign for the workability of competition in the U.S. airline industry, horizontal merger
assessments remain an important part of public policy in the U.S. airline industry. This is
particularly true as our merger sample only includes mergers with largely complementary
networks (which received antitrust clearance beforehand) and other (potentially
anticompetitive) mergers would have faced severe difficulties to receive clearance from the
antitrust authority.22
21 For the group of liquidation-related exits, our analysis reveals that entry activities cannot drive average prices down to the pre-exit level.
22 For example, the European Commission (EC) recently prohibited two mergers which both involved shared hubs: Dublin in case of the Ryanair-Aer Lingus merger proposal (Case No COMP/M.4439, decided in 2007) and Athens in case of the Olympic Air-Aegean Airlines merger proposal (Case No COMP/M.5830, decided in 2011). In both cases, the EC concluded that (route) competition would be harmed substantially by the mergers and therefore prohibited the transactions. In the United States, several merger proposals were abandoned after the DOJ signaled competition concerns. For example, in 2001, United Airlines and US Air ended their merger plans after the DOJ announced its intent to block the transaction (see, e.g., U.S. General Accounting Office, 2001 for an analysis of the expected competitive effects of the proposed merger). Three years earlier, in 1998, a proposal of Northwest Airlines’ to acquire Continental Airlines received similar signals from the DOJ and was subsequently abandoned.
30
5 SUMMARY AND CONCLUSION
In the last decade, the domestic U.S. airline industry has experienced a substantial
consolidation trend. In addition to a number of high level mergers such as American Airlines
– Trans World Airlines (2001), America West – US Airways (2005) and Delta Air Lines –
Northwest Airlines (2009), several smaller carriers such as National Airlines (2002),
Independence Air (2006) and ATA Airlines (2008) had to leave the industry.
Despite this high relevance of firm exits for the recent development of the domestic U.S.
airline industry, empirical evidence on the effects of these consolidations is rare. Studies
focusing on the market impact of liquidations do not exist to the best of our knowledge and
existing studies on the competitive effects of airline mergers almost exclusively stem from the
1980s and focus on the specific case of a largely overlapping route network of the merging
parties (due to a shared hub). However, such a network structure is rather uncommon in recent
mergers and therefore raises the demand for both a new conceptual framework for
investigating firm exits in the airline industry and a corresponding new empirical analysis of
the effects of such firm exits.
Against this background, we study the effects of firm exits on prices, different measures of
quantity and entry in the domestic U.S. airline industry from 1995 to 2010. Applying fixed
effects models we find that liquidation-related exits have, in the short run, a substantial effect
on average yield. Interestingly, the effect is found to be persistent over time. Turning to the
effects on quantity, our analysis reveals that firm exits due to liquidation lead to a large
decrease in capacity and demand. Furthermore, a large and quite persistent entry-inducing
effect is observed shortly after liquidation exits.
The effects of merger-related exits are assessed for two different route types: overlapping
routes and switching routes (i.e., routes which experience a merger-induced switch of the
operating airline). In the short run, estimation results show a significant price increase on both
route types. In the long run, however, it is found that these price increases vanish. Prices even
decrease on switching monopoly routes. Both findings can be interpreted as clear indications
for the realization and pass-on of merger efficiencies. The capacity and demand reductions
following firm exit through merger are generally smaller than those for liquidation-related
exits. Entry is only induced to a small degree in the long run. However, for the sub-sample of
31
small mergers moderate entry takes place relatively early after the completion of the merger.
Finally, it is worth mentioning that our approach to study the competitive effects of firm
exit allows the derivation of several avenues for future research. In addition to possible
changes in the empirical strategy or the application of alternative estimation approaches, a
particularly interesting research area is econometric case studies of the effects of particularly
large mergers. Such investigations would not only allow a much more detailed assessment of
the competitive effects, e.g., through the construction of much more detailed route categories,
but would also enable ex-post evaluations of the respective merger decisions of the antitrust
authority. Such case-study related research is therefore likely to create important positive
spillover effects on the welfare-improving impact of antitrust policy in general and the quality
of future merger assessments in particular.
32
REFERENCES
Allison, P. (2001), Missing Data, Thousand Oaks.
Baumol, W., J. Panzar and R. Willig (1982), Contestable Markets and the Theory of Industry Structure, San Diego.
Benkard, L., A. Bodoh-Creed and J. Lazarev (2010), Simulating the Dynamic Effects of Horizontal Mergers: U.S. Airlines, Working Paper, Stanford University, Palo Alto.
Beutel, P. and M. McBride (1992), Market Power and the Northwest-Republic Airline Merger: A Residual Demand Approach, Southern Economic Journal, 58, 709-720.
Bilotkach, V. (2011), Multimarket Contact and Intensity of Competition: Evidence from an Airline Merger, Review of Industrial Organization, 38, 95-115.
Borenstein, S. (1990), Airline Mergers, Airport Dominance, and Market Power, American Economic Review-Papers and Proceedings, 80, 400-404.
Borenstein, S. and N. Rose (2008), How Airline Markets Work…Or Do They? Regulatory Reform in the Airline Industry, forthcoming in: Rose, N., Economic Regulation and Its Reform: What Have We Learned?, Chicago.
Borenstein, S. and N. Rose (2003), The Impact of Bankruptcy on Airline Service Levels, American Economic Review, Papers and Proceedings, 93, 415-419.
Borenstein, S. and N. Rose (1995), Bankruptcy and Pricing in U.S. Airline Markets, American Economic Review, Papers and Proceedings, 85, 397-402.
Brueckner, J., N. Dyer and P. Spiller (1992), Fare Determination in Airline Hub-and-Spoke Networks, RAND Journal of Economics, 23, 309-333.
Busse, M. (2002), Firm Financial Condition and Airline Price Wars, RAND Journal of Economics, 33, 298-318.
Butler, R. and J. Huston (1989), Merger Mania and Airline Fares, Eastern Economic Journal, 15, 7-16.
Ciliberto, F. and C. Schenone (2008), Bankruptcy and Product-Market Competition: Evidence from the Airline Industry, Working Paper, University of Virginia.
Daraban, B. and G. Fournier (2008), Incumbent Responses to Low-Cost Airline Entry and Exit: A Spatial Autoregressive Panel Data Analysis, Research in Transportation Economics, 24, 15-24.
Geroski, P. A. (1991). Market Dynamics and Entry. Oxford: Oxford University Press.
Geroski, P. A. (1995). What Do We Know About Entry? International Journal of Industrial Organization, 13, pp. 421-440.
Hüschelrath, K. and K. Müller (2011), Patterns and Effects of Entry in U.S. Airline Markets, Journal of Industry, Competition and Trade, forthcoming.
33
Jordan, W. (1988), Problems Stemming from Airline Mergers and Acquisitions, Transportation Journal, 27, 9-30.
Kim, E. and V. Singal (1993), Mergers and Market Power: Evidence from the Airline Industry, American Economic Review, 83, 549-569.
Knapp, W. (1990), Event Analysis of Air Carrier Mergers and Acquisitions, Review of Economics and Statistics, 57, 703-707.
Kwoka, J. and E. Shumilkina (2010), The Price Effect of Eliminating Potential Competition: Evidence from an Airline Merger, Journal of Industrial Economics, 58, 767-793.
McGuckin, R., F. Warren-Boulton and P. Waldstein (1992), The Use of Stock Market Returns in Antitrust Analysis of Mergers, Review of Industrial Organization, 7, 1-11.
Morrison, S. (1996), Airline Mergers - A Longer View, Journal of Transport Economics and Policy, 30, 237-250.
Schumpeter, J. A. (1942). Capitalism, Socialism and Democracy. New York: Harper & Brothers Publishers.
Singal, V. (1996), Airline Mergers and Competition: An Integration of Stock and Product Price Effects, Journal of Business, 69, 233-268.
Slovin, M., M. Sushka and C. Hudson (1991), Deregulation, Contestability, and Airline Acquisitions, Journal of Financial Economics, 30, 231-251.
U.S. General Accounting Office (2001), Airline Competition – Issues Raised by Consolidation Proposals, GAO-01-370T, Washington D.C.
U.S. Government Accountability Office (2010), Airline Mergers – Issues Raised by the Proposed Merger of United and Continental Airlines, GAO-10-778T, Washington D.C.
Werden, G., L. Froeb (1998), The Entry-Inducing Effects of Horizontal Mergers: An Exploratory Analysis, Journal of Industrial Economics, 46, 525-543.
Werden, G., A. Joskow and R. Johnson (1991), The Effects of Mergers on Price and Output: Two Case Studies from the Airline Industry, Managerial and Decision Economics, 12, 341-352.
Williamson, O. (1968), Economies as an Antitrust Defense: The Welfare Tradeoffs, American Economic Review, 58, 18-36.
APPENDIX
Table 9: Summary statistics - operational exits
quarter before/of exit quarter after exit period before exit period after exit mean s.d. mean s.d. mean s.d. mean s.d.
ln(non-stop yield) 2.829 (0.662) 2.859 (0.661) 2.867 (0.665) 2.872 (0.665) non-stop yield 21.471 (17.321) 22.089 (17.532) 22.337 (17.832) 22.508 (18.177) ln(departures) 6.991 (0.812) 6.882 (0.904) 6.990 (0.838) 6.922 (0.875) departures 1,476 (1,236) 1,392 (1,229) 1,494 (1275) 1,430 (1268) ln(Passengers) 11.328 (0.911) 11.203 (0.994) 11.352 (0.917) 11.256 (0.964) Passengers 122,031 (113,940) 112,891 (109,714) 124,630 (114,913) 117,285 (113,523) # airlines w/o exit 0.064 (0.370) 0.283 (0.520) 0.018 (0.306) 0.031 (0.349) post-exit monopoly 0.425 (0.495) 0.425 (0.495) 0.425 (0.494) 0.423 (0.494) # airlines w/o exit 1.541 (0.950) 1.825 (0.910) 1.447 (0.959) 1.817 (0.882) # LCCs w/o exit 0.398 (0.564) 0.404 (0.571) 0.339 (0.526) 0.429 (0.596) avg. plane size 117.156 (42.346) 114.341 (45.406) 121.731 (41.209) 114.559 (45.722) ln(one-stop yield) 2.733 (1.151) 2.798 (1.019) 2.840 (1.006) 2.807 (1.032) missing one-stop yield 0.025 (0.156) 0.016 (0.127) 0.015 (0.122) 0.017 (0.128) airport size 1.843 (0.807) 1.827 (0.800) 1.848 (0.797) 1.831 (0.800) ln(# establ. ) 11.589 (0.743) 11.593 (0.744) 11.573 (0.743) 11.607 (0.745) ln(avg. weekly wage) 6.682 (0.170) 6.695 (0.167) 6.658 (0.175) 6.716 (0.165) ln(labor force) 14.492 (0.685) 14.494 (0.684) 14.482 (0.686) 14.502 (0.684) Year 2004 (3.737) 2004 (3.756) 2003 (3.766) 2005 (3.753) Quarter 2 0.228 (0.420) 0.219 (0.414) 0.251 (0.433) 0.252 (0.434) Quarter 3 0.285 (0.452) 0.228 (0.420) 0.250 (0.433) 0.246 (0.430) Quarter 4 0.268 (0.443) 0.285 (0.452) 0.249 (0.432) 0.249 (0.433) Observations 918 918 7,319 7,271
Notes: Prices in 1995 $ cents. Sources: U.S. DOT, T-100 Domestic Segment Data, Airline Origin and Destination and U.S. Bureau of Labor Statistics,
authors’ calculations.
Table 10: Summary statistics - merger exits (switching routes)
quarter before/of exit quarter after exit period before exit period after exit mean s.d. mean s.d. mean s.d. mean s.d. ln(non-stop yield) 2.594 (0.527) 2.603 (0.512) 2.607 (0.538) 2.609 (0.489) non-stop yield 15.477 (9.338) 15.512 (9.221) 15.820 (9.966) 15.389 (8.364) ln(departures) 7.010 (0.852) 7.025 (0.827) 6.994 (0.830) 7.057 (0.844) departures 1,525 (1,206) 1,532 (1,223) 1,486 (1,179) 1,598 (1,275) ln(Passengers) 11.687 (0.827) 11.660 (0.826) 11.673 (0.815) 11.709 (0.838) Passengers 161,172 (122,384) 157,622 (122,376) 158,124 (120,170) 167,396 (132,690) # airlines w/o exit -0.005 (0.204) 0.005 (0.297) -0.008 (0.274) 0.014 (0.242) post-exit monopoly 0.276 (0.448) 0.276 (0.448) 0.276 (0.447) 0.259 (0.438) # airlines w/o exit 1.005 (0.814) 1.009 (0.844) 1.051 (0.860) 1.083 (0.865) # LCCs w/o exit 0.498 (0.528) 0.493 (0.554) 0.503 (0.550) 0.543 (0.582) avg. plane size 143.387 (23.694) 142.432 (24.771) 144.700 (25.261) 141.724 (24.495) ln(one-stop yield) 2.645 (0.625) 2.658 (0.616) 2.635 (0.705) 2.679 (0.603) missing one-stop yield 0.000 (0.000) 0.000 (0.000) 0.003 (0.059) 0.000 (0.000) airport size 2.107 (0.759) 2.140 (0.770) 2.125 (0.740) 2.141 (0.799) ln(# establ. ) 11.389 (0.583) 11.391 (0.579) 11.369 (0.583) 11.405 (0.576) ln(avg. weekly wage) 6.690 (0.175) 6.682 (0.166) 6.647 (0.172) 6.703 (0.157) ln(labor force) 14.310 (0.515) 14.312 (0.515) 14.297 (0.517) 14.320 (0.513) Year 2004 (3.915) 2004 (4.096) 2003 (4.079) 2005 (3.931) Quarter 2 0.207 (0.406) 0.069 (0.254) 0.250 (0.433) 0.267 (0.443) Quarter 3 0.373 (0.485) 0.207 (0.406) 0.249 (0.433) 0.231 (0.422) Quarter 4 0.350 (0.478) 0.373 (0.485) 0.250 (0.433) 0.233 (0.423) Observations 217 217 1,734 1,620
Notes: Prices in 1995 $ cents. Sources: U.S. DOT, T-100 Domestic Segment Data, Airline Origin and Destination and U.S. Bureau of Labor Statistics,
authors’ calculations.
Table 11: Summary statistics - merger exits (overlapping routes)
quarter before/of exit quarter after exit period before exit period after exit mean s.d. mean s.d. mean s.d. mean s.d. ln(non-stop yield) 3.060 (0.687) 3.121 (0.713) 3.132 (0.689) 3.079 (0.700) non-stop yield 26.741 (18.466) 28.805 (20.195) 28.813 (19.850) 27.600 (19.625) ln(departures) 7.059 (0.671) 7.032 (0.661) 7.061 (0.654) 7.070 (0.658) departures 1,492 (1,338) 1,449 (1,328) 1,480 (1299) 1,506 (1420) ln(Passengers) 11.414 (0.795) 11.309 (0.838) 11.471 (0.771) 11.441 (0.813) Passengers 129,407 (139,133) 121,612 (142,607) 135,024 (141,801) 134,591 (149,189) # airlines w/o exit -0.025 (0.158) 0.038 (0.250) 0.005 (0.222) 0.016 (0.226) post-exit monopoly 0.468 (0.502) 0.468 (0.502) 0.468 (0.499) 0.448 (0.498) # airlines w/o exit 0.772 (1.208) 0.810 (1.220) 0.739 (1.180) 0.890 (1.281) # LCCs w/o exit 0.241 (0.459) 0.278 (0.505) 0.201 (0.413) 0.310 (0.515) avg. plane size 108.889 (38.023) 105.433 (39.557) 114.517 (37.854) 111.790 (41.470) ln(one-stop yield) 2.760 (1.311) 2.973 (0.964) 2.925 (1.079) 2.921 (1.124) missing one-stop yield 0.038 (0.192) 0.013 (0.113) 0.021 (0.142) 0.022 (0.146) airport size 1.745 (0.697) 1.759 (0.688) 1.789 (0.694) 1.781 (0.689) ln(# establ. ) 11.502 (0.672) 11.492 (0.679) 11.503 (0.661) 11.535 (0.690) ln(avg. weekly wage) 6.841 (0.132) 6.800 (0.132) 6.791 (0.128) 6.819 (0.124) ln(labor force) 14.460 (0.600) 14.456 (0.602) 14.458 (0.594) 14.488 (0.609) Year 2008 (2.885) 2008 (3.226) 2007 (3.099) 2009 (3.299) Quarter 2 0.101 (0.304) 0.025 (0.158) 0.250 (0.433) 0.310 (0.463) Quarter 3 0.139 (0.348) 0.101 (0.304) 0.250 (0.433) 0.183 (0.387) Quarter 4 0.734 (0.445) 0.139 (0.348) 0.250 (0.433) 0.196 (0.398) Observations 79 79 632 509
Notes: Prices in 1995 $ cents. Sources: U.S. DOT, T-100 Domestic Segment Data, Airline Origin and Destination and U.S. Bureau of Labor Statistics,
authors’ calculations.
Table 12: Summary statistics – liquidation exits
before/of exit after exit period before exit period after exit mean s.d. mean s.d. mean s.d. mean s.d. ln(non-stop yield) 2.373 (0.525) 2.487 (0.534) 2.420 (0.534) 2.483 (0.547) non-stop yield 12.337 (6.898) 13.997 (8.709) 13.250 (9.549) 14.157 (9.439) ln(departures) 7.263 (0.733) 7.059 (0.841) 7.263 (0.790) 7.133 (0.798) Departures 1,819 (1,323) 1,560 (1,182) 1,900 (1523) 1,639 (1145) ln(Passengers) 11.896 (0.743) 11.699 (0.853) 11.903 (0.773) 11.798 (0.810) Passengers 189,951 (144,791) 167,394 (143,222) 195,816 (157,550) 177,768 (132,856) # airlines w/o exit 0.023 (0.151) 0.205 (0.408) 0.006 (0.262) 0.031 (0.365) post-exit monopoly 0.318 (0.471) 0.318 (0.471) 0.318 (0.466) 0.318 (0.466) # airlines w/o exit 2.023 (1.285) 2.227 (1.309) 2.034 (1.331) 2.230 (1.155) # LCCs w/o exit 0.295 (0.462) 0.409 (0.497) 0.276 (0.447) 0.517 (0.594) avg. plane size 146.421 (42.781) 147.572 (46.584) 147.936 (43.786) 149.001 (44.010) ln(one-stop yield) 2.492 (0.678) 2.583 (0.668) 2.532 (0.659) 2.536 (0.660) missing one-stop yield 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) 0.000 (0.000) airport size 2.057 (0.735) 1.945 (0.715) 2.051 (0.722) 1.996 (0.725) ln(# establ. ) 11.809 (0.619) 11.815 (0.618) 11.788 (0.620) 11.832 (0.614) ln(avg. weekly wage) 6.746 (0.180) 6.764 (0.147) 6.710 (0.147) 6.771 (0.152) ln(labor force) 14.684 (0.578) 14.689 (0.577) 14.672 (0.571) 14.695 (0.569) Year 2004 (2.548) 2004 (2.726) 2003 (2.566) 2005 (2.566) Quarter 2 0.000 (0.000) 0.273 (0.451) 0.250 (0.434) 0.250 (0.434) Quarter 3 0.432 (0.501) 0.000 (0.000) 0.250 (0.434) 0.250 (0.434) Quarter 4 0.295 (0.462) 0.432 (0.501) 0.250 (0.434) 0.250 (0.434) Observations 44 44 352 352
Notes: Prices in 1995 $ cents. Sources: U.S. DOT, T-100 Domestic Segment Data, Airline Origin and Destination and U.S. Bureau of Labor Statistics,
authors’ calculations.
Table 13: Fixed effects regressions for the effect of exits on non-stop yield (excluding large mergers)
ln(non-stop yield) – short term ln(non-stop yield) – medium term ln(non-stop yield) – long termVariable coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.)
operational exit 0.065 *** (0.006) 0.057*** (0.007) 0.071*** (0.006) 0.060*** (0.007) 0.060*** (0.006) 0.050*** (0.007) merger exit (switching) 0.003 (0.022) 0.000 (0.023) 0.003 (0.022) -0.003 (0.023) 0.011 (0.025) 0.006 (0.027) merger exit (overlap) -0.086 (0.054) -0.087 (0.054) -0.066 (0.054) -0.067 (0.054) -0.066 (0.056) -0.067 (0.056) liquidation exit 0.125 *** (0.017) 0.111*** (0.020) 0.129*** (0.017) 0.117*** (0.018) 0.113*** (0.017) 0.105*** (0.019) Op. exit # monopoly 0.018* (0.010) 0.025** (0.010) 0.023** (0.010) m. ex. (sw.) # monopoly 0.033 (0.055) 0.075 (0.062) 0.061 (0.052) m. ex. (ov.) # monopoly - - - liquid. exit # monopoly 0.042 (0.035) 0.039 (0.037) 0.023 (0.036) # airlines w/o exit -0.019 *** (0.006) -0.018*** (0.006) -0.022*** (0.006) -0.019*** (0.006) -0.035*** (0.005) -0.034*** (0.005) # LCCs w/o exit -0.103 *** (0.016) -0.102*** (0.016) -0.104*** (0.014) -0.103*** (0.014) -0.094*** (0.013) -0.093*** (0.013) Avg. plane size -0.001 *** (0.000) -0.001*** (0.000) -0.000** (0.000) -0.000** (0.000) -0.000* (0.000) -0.000* (0.000) ln(one-stop yield) 0.355 *** (0.029) 0.354*** (0.029) 0.383*** (0.027) 0.382*** (0.027) 0.368*** (0.033) 0.369*** (0.033) missing one-stop yield 2.439 *** (0.204) 2.434*** (0.205) 2.642*** (0.194) 2.639*** (0.194) 2.533*** (0.236) 2.532*** (0.237) airport size (mean) -0.042 * (0.022) -0.042* (0.022) -0.036* (0.021) -0.036* (0.021) -0.028 (0.020) -0.028 (0.020) ln(# establ. ) (mean) 0.452 *** (0.119) 0.461*** (0.119) 0.337*** (0.109) 0.359*** (0.110) 0.118 (0.096) 0.139 (0.097) ln(avg. weekly wage) 0.171 *** (0.066) 0.172*** (0.066) 0.178*** (0.057) 0.180*** (0.057) 0.190*** (0.058) 0.190*** (0.058) ln(labor force) 0.615 ** (0.250) 0.621** (0.250) 0.605** (0.241) 0.600** (0.241) 0.676*** (0.208) 0.664*** (0.208) Year -0.057 *** (0.006) -0.058*** (0.006) -0.048*** (0.005) -0.049*** (0.005) -0.027*** (0.004) -0.028*** (0.004) Quarter 2 -0.016 *** (0.004) -0.016*** (0.004) -0.015*** (0.004) -0.015*** (0.004) -0.007** (0.003) -0.007** (0.003) Quarter 3 -0.043 *** (0.006) -0.043*** (0.006) -0.038*** (0.005) -0.038*** (0.005) -0.025*** (0.004) -0.025*** (0.004) Quarter 4 -0.068 *** (0.007) -0.068*** (0.007) -0.060*** (0.006) -0.061*** (0.006) -0.045*** (0.005) -0.045*** (0.005) Constant 101.280 *** (10.025) 101.959*** (10.046) 84.567*** (8.546) 85.556*** (8.562) 44.498*** (6.281) 45.015*** (6.280) R2 0.264/0.108/0.109 0.265/0.105/0.105 0.274/0.150/0.150 0.276/0.144/0.144 0.254/0.186/0.185 0.256/0.181/0.180 Observations 8,875 8,875 10,655 10,655 14,152 14,152 Routes 890 890 890 890 890 890
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data, Airline Origin and Destination Survey (DB1B) and U.S. Bureau of Labor Statistics, authors’ calculations.
Table 14: Fixed effects regressions for the effect of exits on number of departures (excluding large mergers)
ln(departures) – short term ln(departures) – medium term ln(departures) – long termVariables coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) operational exit -0.216 *** (0.013) -0.133*** (0.013) -0.209*** (0.013) -0.131*** (0.013) -0.210*** (0.013) -0.142*** (0.013) merger exit (switching) -0.089 *** (0.032) -0.072** (0.031) -0.060* (0.031) -0.062* (0.032) -0.077** (0.037) -0.093** (0.038) merger exit (overlap) -0.044 (0.045) -0.036 (0.045) -0.036 (0.055) -0.029 (0.056) -0.044 (0.065) -0.039 (0.065) liquidation exit -0.173 *** (0.033) -0.143*** (0.036) -0.175*** (0.029) -0.148*** (0.035) -0.186*** (0.031) -0.169*** (0.039) op. exit # monopoly -0.186*** (0.027) -0.175*** (0.024) -0.157*** (0.022) m. exit (sw.)#monopoly -0.193 (0.130) 0.039 (0.072) 0.212** (0.095) m. exit (ov.) # monopoly liq. exit # monopoly -0.082 (0.070) -0.076 (0.055) -0.037 (0.053) # airlines w/o exit 0.158 *** (0.015) 0.140*** (0.015) 0.155*** (0.015) 0.139*** (0.015) 0.146*** (0.015) 0.137*** (0.015) # LCCs w/o exit 0.058 ** (0.029) 0.052* (0.029) 0.084*** (0.026) 0.075*** (0.026) 0.094*** (0.023) 0.085*** (0.023) Avg. plane size -0.004 *** (0.001) -0.004*** (0.001) -0.004*** (0.001) -0.004*** (0.001) -0.005*** (0.001) -0.005*** (0.001) airport size (mean) 0.769 *** (0.051) 0.773*** (0.049) 0.781*** (0.053) 0.787*** (0.050) 0.791*** (0.052) 0.795*** (0.049) ln(# establ. ) (mean) 0.433 ** (0.218) 0.308 (0.212) 0.675*** (0.209) 0.511** (0.201) 0.690*** (0.188) 0.558*** (0.184) ln(avg. weekly wage) (mean) 0.718 *** (0.128) 0.720*** (0.125) 0.747*** (0.120) 0.756*** (0.119) 0.714*** (0.109) 0.737*** (0.108) ln(labor force) (mean) 0.906 * (0.474) 0.891* (0.468) 0.523 (0.454) 0.561 (0.453) 0.159 (0.399) 0.211 (0.397) Year -0.021 ** (0.010) -0.018* (0.010) -0.020** (0.009) -0.016* (0.009) -0.012* (0.007) -0.011 (0.007) Quarter 2 0.068 *** (0.009) 0.069*** (0.009) 0.065*** (0.009) 0.066*** (0.008) 0.065*** (0.007) 0.067*** (0.007) Quarter 3 0.068 *** (0.011) 0.070*** (0.011) 0.065*** (0.010) 0.067*** (0.010) 0.062*** (0.009) 0.064*** (0.009) Quarter 4 -0.014 (0.010) -0.011 (0.010) -0.021** (0.009) -0.018* (0.009) -0.019** (0.008) -0.017** (0.008) Constant 25.667 (16.458) 19.999 (16.467) 24.827 (15.124) 19.125 (15.019) 15.670 (12.391) 13.304 (12.270) R2 (within/between/overall) 0.220/0.203/0.194 0.237/0.215/0.207 0.239/0.221/0.211 0.259/0.236/0.227 0.257/0.264/0.254 0.275/0.283/0.272 Observations 8,875 8,875 10,655 10,655 14,152 14,152 Routes 890 890 890 890 890 890
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data and U.S. Bureau of Labor Statistics, authors’ calculations.
40
Table 15: Fixed effects regressions for the effect of exits on number of passengers (excluding large mergers)
ln(passengers) – short term ln(passengers) – medium term ln(passengers) – long termVariables coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) coef. (s.e.) operational exit -0.186 *** (0.012) -0.104*** (0.013) -0.181*** (0.013) -0.102*** (0.013) -0.187*** (0.013) -0.117*** (0.013) merger exit (switching) -0.091 *** (0.029) -0.073*** (0.028) -0.060** (0.028) -0.064** (0.029) -0.075** (0.034) -0.100*** (0.033) merger exit (overlap) -0.053 (0.041) -0.044 (0.041) -0.013 (0.053) -0.006 (0.053) -0.035 (0.060) -0.030 (0.060) liquidation exit -0.150 *** (0.040) -0.075** (0.037) -0.161*** (0.039) -0.091** (0.037) -0.182*** (0.036) -0.131*** (0.040) op. exit # monopoly -0.183*** (0.027) -0.177*** (0.025) -0.160*** (0.022) m. ex. (sw.) # monopoly -0.202* (0.111) 0.067 (0.093) 0.338*** (0.091) m. ex. (ov.) # monopoly liq. exit # monopoly -0.223** (0.089) -0.208** (0.083) -0.147** (0.067) # airlines w/o exit 0.139 *** (0.014) 0.122*** (0.014) 0.136*** (0.014) 0.121*** (0.014) 0.130*** (0.014) 0.122*** (0.013) # LCCs w/o exit 0.084 *** (0.030) 0.076** (0.030) 0.111*** (0.027) 0.100*** (0.026) 0.119*** (0.023) 0.108*** (0.023) avg. plane size 0.005 *** (0.001) 0.005*** (0.001) 0.004*** (0.001) 0.004*** (0.001) 0.004*** (0.001) 0.004*** (0.001) airport size (mean) 1.014 *** (0.054) 1.018*** (0.052) 1.003*** (0.057) 1.008*** (0.054) 0.981*** (0.054) 0.985*** (0.051) ln(# establ. ) (mean) 0.272 (0.214) 0.169 (0.205) 0.544*** (0.206) 0.398** (0.195) 0.605*** (0.187) 0.484*** (0.180) ln(avg. weekly wage) (mean) 0.560 *** (0.131) 0.557*** (0.129) 0.649*** (0.123) 0.652*** (0.123) 0.695*** (0.111) 0.714*** (0.112) ln(labor force) (mean) 1.080 ** (0.474) 1.047** (0.465) 0.595 (0.451) 0.617 (0.444) 0.180 (0.393) 0.219 (0.384) Year -0.007 (0.010) -0.003 (0.010) -0.007 (0.009) -0.003 (0.009) -0.000 (0.007) 0.002 (0.007) Quarter 2 0.152 *** (0.009) 0.153*** (0.009) 0.149*** (0.009) 0.150*** (0.009) 0.153*** (0.007) 0.155*** (0.007) Quarter 3 0.137 *** (0.012) 0.139*** (0.012) 0.136*** (0.010) 0.138*** (0.010) 0.138*** (0.009) 0.140*** (0.009) Quarter 4 0.026 ** (0.011) 0.029*** (0.011) 0.016 (0.010) 0.020** (0.010) 0.017* (0.009) 0.019** (0.009) Constant 0.129 (16.776) -5.666 (16.807) 3.813 (15.627) -2.274 (15.571) -5.273 (12.860) -7.886 (12.813) R2 (within/between/overall) 0.271/0.270/0.260 0.287/0.283/0.273 0.272/0.300/0.288 0.292/0.315/0.303 0.272/0.348/0.333 0.290/0.363/0.349 Observations 8,875 8,875 10,655 10,655 14,152 14,152 Routes 890 890 890 890 890 890
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data and U.S. Bureau of Labor Statistics, authors’ calculations.
41
Table 16: Fixed effects regressions for the effects of exits on entry (excluding large mergers)
airlines w/o exiting/merger - short term
airlines w/o exiting/merger – medium term
airlines w/o exiting/merger - long term
coef. (s.e.) coef. (s.e.) coef. (s.e.) operational exit 0.204 *** (0.015) 0.167*** (0.014) 0.170 *** (0.013) merger exit (switching) 0.058 (0.039) 0.063** (0.029) 0.110 *** (0.027) merger exit (overlap) 0.182 ** (0.088) 0.150** (0.065) 0.184 *** (0.048) liquidation exit 0.125 *** (0.043) 0.111** (0.044) 0.157 *** (0.040) # airlines w/o exit (lag) -0.390 *** (0.019) -0.350*** (0.016) -0.303 *** (0.013) # LCCs w/o exit (lag) 0.027 (0.031) 0.029 (0.026) 0.018 (0.020) avg. plane size -0.004 *** (0.001) -0.004*** (0.000) -0.004 *** (0.000) airport size (mean) 0.289 *** (0.046) 0.264*** (0.041) 0.258 *** (0.033) ln(# establ. ) (mean) -0.371 (0.312) -0.281 (0.250) -0.243 (0.161) ln(avg. weekly wage) (mean) 0.013 (0.155) -0.125 (0.121) -0.066 (0.095) ln(labor force) (mean) -2.197 *** (0.521) -1.768*** (0.424) -1.244 *** (0.295) Year 0.044 *** (0.011) 0.020** (0.009) -0.011 * (0.006) Quarter 2 0.011 (0.012) 0.001 (0.010) 0.006 (0.008) Quarter 3 -0.015 (0.012) -0.032*** (0.010) -0.031 *** (0.008) Quarter 4 0.038 *** (0.014) 0.027** (0.012) 0.008 (0.010) Constant -51.711 *** (18.160) -8.963 (15.949) 44.411 *** (10.974) R2 (within/between/overall) 0.215 / 0.002 / 0.002 0.192 / 0.002 / 0.003 0.172 / 0.002 / 0.004 Observations 8,838 10,618 14,115 Routes 890 890 890
Notes: Significance levels *** p<0.01, ** p<0.05, * p<0.1, cluster-robust standard errors in parentheses. Sources: U.S. DOT, T-100 Domestic Segment Data and U.S. Bureau of Labor Statistics, authors’ calculations.